Heterogeneity of CEO Social Networks and Firm Value

Transcription

Yiwei Fang
Stuart School of Business
Illinois Institute of Technology
565 W. Adams Street, 4th Floor
Chicago, IL 60661
Telephone: 312 906 6559
E-mail: [email protected]
Bill Francis
Lally School of Management
Rensselaer Polytechnic Institute
110 8th Street, Pittsburgh Building
Troy, New York 12180-3590
Telephone: 518 276 3908
Iftekhar Hasan
Fordham University and Bank of Finland
5 Columbus Circle, 11th Floor
New York, New York 10019
Telephone: 646 312 8278
1
Abstract
This paper examines through various channels the effects of heterogeneity in CEO social
networks on firm value. We construct four measures of heterogeneity based on demographic
attributes, intellectual backgrounds, professional experience, and international exposures of
individuals in the social network. We find that heterogeneity in CEO social networks increases
Tobin's Q. We also find that greater heterogeneity in CEO social networks leads to more
innovation, more foreign sales growth, higher investment sensitivity to Tobin’s Q, and better
M&A performance. Overall, our results indicate that the heterogeneity of CEO social networks is
an aspect of CEO social capital and soft skills that deserves the attention of shareholders.
Keywords: Social Networks, Heterogeneity, Intellectual Background, Professional Experience,
International Exposure and Firm Value.
JEL code: L14; J15; J24; G30;
2
1. Introduction
The United States is a melting pot of different races and cultures. Since the 1970s, the percentage
of the population that is foreign-born has increased significantly due to higher fertility rates and
higher labor-based immigration rates (Shrestha, 2011). The projected demographic
characteristics of the U.S. workforce suggest that immigration trends will continue to bring more
diversity to the American workforce. By 2050, minorities are projected to rise from one in every
four Americans to almost one in every two. The gender gap in the workforce will also narrow
(Toossi, 2002).
The increasing diversity of the workforce necessitates a better understanding of the
economic influence of individual differences (Pfeffer, 1983; Ashraf and Galor, 2013; Fryer and
Loury, 2013). In turn, one important insight from economic studies is that variety within human
populations gives rise to knowledge heterogeneity, which is crucial to the production of
innovation and the accumulation of universally applicable human capital (Hargadon and Sutton,
1997; Galunic and Rodan, 1998). Such an observation has an evidential basis in human biology,
as genetic variations are associated with different modes of cognitive functioning, approaches,
and ideologies. Hence, access to a more diverse group of people could create a wider range of
options, reduce groupthink, and make one knowledgeable of and adaptable to a rapidly changing
environment (Asch, 1951; Janis, 1982). However, a high degree of diversity might also hinder
cooperation and trust between individuals, lead to higher turnover rates among group members,
and therefore be detrimental to value creation (Wagner et al., 1984; Pelled, 1996; Ashraf and
Galor, 2013).
In this paper, we build upon this line of research and study the impacts of socialenvironmental diversity on firm value through the lens of CEO social networks. In particular, we
ask the following questions: Does shareholder wealth rise if a CEO is connected with a more
heterogeneous group of people who have diverse experience and backgrounds? And if so,
through which channel(s) does this heterogeneity make manifest itself?
3
We focus on social networks because they can influence decision-making in significant
ways. First of all, they facilitate information diffusion and transmission within the group (Holzer,
1987; Granovetter, 1995; Calvo-Armengol and Jackson, 2004), and therefore managers often
draw on their social networks to obtain valuable experiences, gather key market information,
exchange resources, and identify business opportunities (Engelberg et al., 2013). Second, when
individuals do not have all the required information, they tend to rely on whatever information
they can acquire via word-of-mouth communication (Ellison and Fudenberg, 1993, 1995; Watts,
2003). Managers likely have similar preferences regarding financial decisions due to the actions
of their social peers (Fracassi, 2012).
Although the role of CEO social networks has been studied in various finance contexts
(e.g., Cohen et al., 2008; Engelberg et al., 2012), the diversity characteristics of social networks
are overlooked. To measure the heterogeneity in CEO social networks (CEO SNH hereafter), we
obtain biographical information for 3,506 CEOs from 2,484 U.S. public firms over the period
2000–2010. We identify their education ties, previous work ties, and other social ties through
activities in charities and country clubs. The generally accepted definition of diversity refers to
differences between individuals on any attributes that could lead to an individual’s perception
that another person is different from himself/herself (Williams and O'Reilly, 1998). In this
regard, our paper addresses more than age, ethnic, and gender diversity; it considers differences
in skills, competencies, and international experiences as well.
Our baseline estimations examine the effect of CEO SNH on firm value using OLS
regressions. Controlling for social-network size, centrality (measure of connectedness), and
various firm and CEO characteristics, we find that greater diversity in terms of intellectual
background, professional experience, and international exposure is significantly associated with
higher firm value. Among these diversity measures, international exposure has the largest impact
on firm value (a $96 million increase in market value following a 10% increase in international
exposure), followed by professional heterogeneity (a $30 million increase in market value
following a 10% increase in professional diversity), and intellectual heterogeneity (a $28 million
increase following a 10% increase in intellectual diversity). Finally, a 10% increase in the overall
4
heterogeneity index (the average of the four individual heterogeneity indices) could increase a
firm’s market value by $106 million. Demographic diversity, however, does not appear to have a
significant impact on firm value.
Though suggestive, the evidence so far is difficult to interpret as a causal relationship
between CEO SNH and firm value. Firms with better performance may have the advantage of
hiring managers who have diverse networks of connections. Moreover, managers of growth
firms have more opportunities to meet people from different backgrounds and thus form diverse
social networks. To deal with the endogeneity of connections, we use death and retirements of
network-member CEOs as the instrument for network ties following Fracassi and Tate (2012).
We find robust results with the instrumental-variable approach. Noticeably, the effect of
demographic diversity on firm value also becomes statistically significant.
We further address the endogeneity problem in an event-study framework to analyze
market reactions to new CEO appointments. We form two groups of CEOs, with the benchmark
group consisting of CEOs whose SNH scores are higher than those of the previous CEOs, and
the matching group consisting of CEOs whose SNH scores are the same or lower than those of
the previous CEOs. We employ the nearest-neighbor matching technique to ensure the two
groups of firms have similar characteristics and CEO personal attributes but different social
networks. The results show that the benchmark firms on average experience significantly
positive cumulative abnormal returns around the time of the CEO appointment. In contrast,
matching firms have insignificant market reactions to the CEO appointment. The results are
robust to using different event windows.
Our analyses further suggest that CEO SNH increases firm value through several
channels. First, we find that innovation activities increase with CEO SNH. When we
simultaneously control for this channel, we find that CEO SNH has less impact on firm value.
This suggests that CEO SNH has a positive impact on innovative capability, and that by acting
on innovation, it adds value to the firm.
Second, we find that firms experience greater foreign-sale growth rates when their CEOs’
networks are more heterogeneous. After simultaneously controlling for the effect of CEO SNH
5
on foreign sales growth, the impact of CEO SNH on firm value becomes less significant. The
results indicate that heterogeneity of CEO social networks has a positive impact on a foreign
sales generation, and that by acting on foreign sales, these CEO social networks add value to
firms.
Finally, we explore the potential channel through investment based on two types of
analysis. We first examine the relationship between CEO SNH and a corporate investment’s
sensitivity to stock price, which becomes a proxy for the efficiency of the corporate investment
(e.g., Dittmar and Shivdasani, 2003; Chen et al., 2007; Xuan, 2009). The results show that
investments are indeed more sensitive to Tobin’s Q when CEOs have more heterogeneous social
ties. Second, we look at M&A investment performance and find a positive relationship between
CEO SNH and five-day cumulative abnormal returns (CAR). The relationship is more
pronounced for diversified deals. In terms of long-run performance, we find that CEO SNH is
associated with better long-run post M&A performance and is more pronounced for diversified
deals.
These findings suggest that diverse CEO social networks provide different knowledge
bases and varieties of opinions that are crucial for helping CEOs make better investment
decisions and achieve more innovation. Moreover, exposures to different cultures also largely
increase CEOs’ ability to establish a network of foreign contacts, identify good business
opportunities, and enjoy more success in foreign business (Prahalad and Hamel, 1990; Reuber
and Fischer, 1997; Carpenter and Sanders, 1998).
Our paper contributes to several different research streams. First, a growing line of
research in behavioral finance tries to understand to the role of CEOs in value creation. For
example, using manager fixed effects as proxies for CEO management style, Bertrand and
Schoar (2003) demonstrate that managers’ personal attributes explain a wide range of corporate
behavior, such as investment, leverage, and cash holdings. Kaplan, Klebanov, and Sorensen
(2012) find that general ability and execution skills are more crucial to firm performance.
Exploring a unique data set, Bennedsen et al. (2011) find that firms whose CEOs are hospitalized
underperform compared to similar firms, but the hospitalization of other senior executives does
6
not have similar effects on firm performance. Along these lines, our paper provides a
microsociology perspective of CEO characteristics. Our results suggest that heterogeneous social
ties should be a key consideration in evaluating CEO quality. CEOs can benefit from diverse
social connections because they provide a broader field of knowledge in response to innovations
in market conditions, international business opportunities, and mergers and acquisitions.
Second, a strand of social-network studies investigate the impact of CEO social networks
on investment style (Fracassi, 2012), R&D (Faleye et al., 2013), M&A (Cai and Sevilir, 2012),
CEO compensation (Barnea and Guedj, 2007; Engelberg et al., 2013; Butler and Gurun, 2012),
mutual fund performance and trading behavior (Cohen et al., 2008; Hong et al., 2005), cost of
capital (Engelberg et al., 2012), analyst performance (Cohen et al., 2010; Horton and Serafeim,
2009), and corporate governance (Hwang and Kim, 2009; Fracassi and Tate, 2012). Though
these studies establish the important economic underpinnings of social networks in various
corporate finance issues, researchers focus primarily on the overall connectedness of CEO social
networks or social ties with resource holders (e.g., banks). Our paper extends the literature by
looking at the heterogeneity characteristics of social ties. By controlling for the size of social
network and potential endogeneity problems, we demonstrate that the value of social networks
also comes from the heterogeneity of their ties.
Last but not least, we identify multiple areas where heterogeneity of CEO social networks
is crucial to add value. The results on innovation suggest that heterogeneous social networks
make CEOs more supportive and successful in spurring product innovations. The findings on
foreign business channels and investment channels show that diversity of social ties enables
CEOs to achieve higher foreign sales and make better investment decisions.
To the best of our knowledge, our paper is the first to establish a direct link between CEO
social-network heterogeneity and firm value. Our findings suggest that with more heterogeneous
social connections, CEOs could obtain broader knowledge, consider a larger range of options,
and learn about more business opportunities. These benefits also enable CEOs to make better
investment decisions and respond more quickly to product innovation and foreign market
competition. Given the changing demographic characteristics of corporate hierarchies, and given
7
the increasing pressure on U.S. companies from global market competition and product
innovation, this study offers important policy implications.
The rest of the paper is organized as follows. Section 2 introduces related literature and
our hypotheses. Section 3 discusses the data and the construction of the variables. Section 4
presents our empirical results. Section 5 summarizes and concludes.
2. Literature Review and Hypotheses
2.1. Social networks and firm value
Fundamentally, social networks affect economic outcomes by offering a way to exchange
resources and information that are not easily accessible elsewhere. Early work on social
networks extensively examines their role in information dissemination in the labor market (e.g.,
spreading news about job vacancies or accounts of workers’ abilities) (Holzer, 1987;
Granovetter, 1995; Calvo-Armengol and Jackson, 2004).
Recent research in finance suggests that social networks also play an influential role in
affecting prices in the financial markets. For example, Cohen et al. (2008) find that portfolio
managers place larger bets and obtain higher returns on firms with which they have social
connections. Similarly, Cohen et al. (2010) and Horton and Serafeim (2009) provide evidence
that analysts provide better forecasts and recommendations when they have educational links to
the company. Cai and Sevilir (2012) show that board connections between targets and acquirers
improve information flow and communication between the two firms, and therefore they benefit
acquirers via lower takeover premiums and greater value creation. Engelberg et al. (2012) report
that interpersonal connections also reduce information asymmetry in the bank loan market. As a
result, firms that have social connections with bankers obtain cheaper bank loans. These findings
support the information role of social networks.
A second way social networks affect economic outcomes is by influencing individual
behavior and strategies. That is, when individuals do not have all the required information, they
tend to rely on whatever information they can acquire via word-of-mouth communication
8
(Ellison and Fudenberg, 1993, 1995; Watts, 2003). Individuals are also likely to have similar
preferences and beliefs due to the actions of their social peers.
Supporting this theory, Fracassi (2012) finds that CEOs who are well connected make
financial decisions similar to those of their social peers. Hong et al. (2005) show that trades of
mutual funds that are in the same city are correlated due to word-of-mouth communication.
Finally, social networks provide social support for connected individuals (Powell, 1990;
McPherson et al., 2001). That is, socially interconnected individuals have more trust in one
another, they tend to interpret one another’s behavior favorably, and they assume that each party
will take predictable and mutually acceptable action (Uzzi, 1996, 1999).
In addition to overall connectedness, the heterogeneity characteristics of social networks
are also crucial determinants of economic outcomes. For example, diverse social networks can
offer a depth and breadth of insight, perspectives, philosophies, and life experiences. This
argument has a root in human biology, which says that genetic variations are associated with
different modes of cognitive functioning, approaches, and ideologies. Therefore, CEOs who are
exposed to people with diverse demographic backgrounds are able to consult with a variety of
authorities, obtain divergent ideas to tackle the same problem, and thus make their companies
more adaptable to market competition. Moreover, different educational backgrounds and
professional experiences help enrich managers’ information sets and provide more approaches to
solve problems (Robinson and Dechant, 1997).
We construct four measures of CEO SNH based on demographic attributes, intellectual
backgrounds, professional experience, and international exposure for individuals in CEO social
networks. We hypothesize that managers derive valuable knowledge from social peers, which
enables them to perform better. Hence, higher heterogeneity of a CEO social network should
lead to higher firm value.
H1: CEO social-network heterogeneity enhances firm value.
2.2. CEO SNH and firm value–innovation channel
9
The literature extensively examines the role of informal networks on innovation. Goyal and
Moraga-Gonzales (2001) demonstrate that R&D networks among firms play an important role in
the knowledge diffusion that promotes innovation. Gomes-Gasseres et al. (2006) find that interorganizational linkages, such as alliances, speed up a firm’s ability to learn and utilize new
technology and innovation (Gomes-Gasseres et al., 2006).
The social psychology approach emphasizes that differences in psychological attributes
and personal characteristics, such as attitudes, cognitive functioning, and beliefs, are also crucial
to achieving a higher level of creativity within a team (Barron and Harrington, 1981; Hong and
Page, 2001; De Dreu and West, 2001). The reason is that diverse individuals bring diverse
opinions and combine ideas that may stimulate innovative solutions to work-related problems.
Hence, knowledge heterogeneity is a wellspring for creativity and innovation (Hargadon and
Sutton, 1997; Galunic and Rodan, 1998).
Supporting this view, Rodan and Galunic (2004) find that managers with access to
heterogeneous knowledge through their social contacts perform better in generating and
implementing new ideas for their firms. Similarly, managers working with a heterogeneous
contextual environment have advantages in developing novel ideas about organizational
products, practices, services, or procedures (Shalley and Gilson, 2004; Shalley and Perry-Smith,
2008). In aligning with these arguments, we propose that innovation is a channel through which
CEO SNH increases firm value.
H2: Heterogeneous CEO social networks have a positive impact on a firm’s innovative
capability, and by acting on innovation, heterogeneity adds value to the firm.
2.3. CEO SNH and firm value–foreign sale channel
The international business literature highlights that cultural sensitivity among corporate leaders
can be critical in the process of internationalization (Reuber and Fischer, 1997). As more and
more firms attempt to penetrate foreign markets, there is an increasing need to match the
diversity of a company’s hierarchy to the diversity of the company’s potential customers
(Robinson and Dechant, 1997). In the finance literature, Carter et al. (2003) find that the
10
ethnicity diversity of boards of directors could make CEOs more sensitive to cultures outside the
U.S. (Carter et al., 2003). Masulis et al. (2012) show that foreign directors can enhance the
advisory capability of boards to the extent that living or working in foreign countries gives them
firsthand knowledge of foreign markets and enables them to develop and tap into a network of
foreign contacts. Foreign directors and their connections can provide valuable assistance to U.S.
corporations, especially those with major foreign operations or aspirations to expand
internationally (Adams et al., 2010).
Hence, we expect that CEOs who have exposure to demographically and intellectually
diverse groups of people are likely to understand different institutional environments and acquire
key foreign-market information. Their connections with foreign companies could also enhance
their managerial capabilities to push sales and generate more foreign business. Given that more
U.S. corporations rely on foreign businesses and that foreign earnings have higher associations
with market valuations than domestic earnings (Bodnar and Weintrop, 1997), we argue that
social network heterogeneity promotes foreign business, pushes sales, and improves firm
performance. This leads to our third hypothesis:
H3: Heterogeneity of CEO social networks has a positive impact on a firm’s foreign sales
generation, and by acting on foreign sales, heterogeneity adds value to the firm.
2.4. CEO SNH and firm value–investment channel
Engelberg et al. (2013) indicate that social networks serve as an informal medium through which
managers share valuable experiences, gather key market information, exchange resources, and
identify business opportunities. It suggests that manager social networks are crucial for
investment performance. Burt (1992) argues people who can build a bridge between two distinct
groups of people will obtain superior information and resources beyond their own groups. It
emphasizes that heterogeneous social connections are also important for decision-makers in
order to consider more alternatives and to take a broader view.
The finance literature examines diversity mainly in the context of boards of directors. For
example, Anderson et al. (2011) investigate the economic value of having a heterogeneous board
11
of directors in a firm. They find that investors value heterogeneous boards in complex firms but
do not value them in less complex firms. Gul et al. (2011) find that gender-diverse boards
improve informativeness by providing more transparent public disclosure in large firms and by
encouraging private information collection in small firms. Additionally, Adams and Ferreira
(2009) show that female board members attend board meetings and interact with management to
a greater degree than male board members. Because of their work ethic and professionalism, the
presence of women on boards in turn has a positive influence on corporate governance. Francis
et al. (2014) find that female CFOs are more risk-averse and tend to make conservative financial
decisions. Francis et al. (2013) examine professional backgrounds of boards of directors, and
their findings suggest that directors who have academic backgrounds can bring to bear their
advising and monitoring expertise, which leads to higher firm performance. Following these
arguments, one of the major benefits of having diverse team members is that doing so generates
healthy debate, adds information richness, and improves corporate governance. These are
important features for decision-making. We therefore expect that diverse social ties are
beneficial for CEOs to make investment decisions.
We examine two investment performance measures: (1) investment-Q sensitivity and (2)
M&A market valuation effect in both short and long run. If a heterogeneous social network helps
managers make better decisions, then, holding all else equal, we expect that:
H4: Heterogeneous CEO social networks lead to more efficient allocation of capital, as
measured by higher investment-Q sensitivity.
H5: Heterogeneity of CEO social networks has a positive impact on a firm’s M&A
performance.
3. Data and Variables
3.1. Measuring CEO social network
CEO social network data is obtained from the BoardEx database provided by Management
Diagnostics Limited. This database contains comprehensive biographic information of senior
12
management and boards of directors of U.S. and European public companies. Biographical
information includes demographic attributes (e.g., date of birth, date of death, gender,
nationality), employment history (e.g., workplaces and job title), educational background (e.g.,
school, degree, and major), and other social activities, such as club membership, professional
associations, and charities. The BoardEx dataset also provide relational links between these
individuals. Links are constructed if two individuals were once employed by the same company
(work ties), graduated from the same university (school ties), or maintained memberships with
the same country clubs and nonprofit organizations, such as charities, government organizations,
and branches of the military (other social ties).
Recent studies that use this dataset tend to focus on one or all three types of social links
(Cohen et al., 2008; Liu, 2010; Engelberg et al., 2012, 2013; Fracassi, 2012; Fracassi and Tate,
2012). In this paper we consider all types of social links. Specifically, we identify a school tie
between two individuals if they went to the same school and graduated within three year of each
other with the same master’s or doctoral degree. The restriction on graduation year and degree is
to maximize the probability that the individuals actually met as a result of shared education
(Fracassi, 2012). Work ties are built if two individuals have worked in the same company. Other
social ties are identified if two individuals maintain membership in the same country clubs, or
served the same charity, university, government, army, or other non-profit association. To make
sure people have active roles in these organizations, we consider a qualified tie to be one in
which a person was more than merely a member and instead maintained an important role in the
organization (with the exception of club membership, for which membership alone is considered
qualified) (Fracassi, 2012). For example, important roles can be “Trustee,” “President,”
“Advisor,” and “Board Member.”
We further refine our criteria in examining CEO social networks by checking the starting
and ending dates of the social relationships. We drop work ties and other social ties that
terminate five years before our testing year. As for school ties, we do not use a time window;
although many CEOs graduated several years ago, they often maintain connections with their
former classmates through alumni events. As regard to work and other social ties, however, this
13
is less likely to be the case. Moreover, we also examine the job titles of the individuals who are
in a CEO’s social network. To maximize the probability that two individuals have actually met
and have had informative conversations, we require that the two individuals must have held
relatively high level positions, such as serving on the top level management team, the board of
directors, an advising council, and so on (Fracassi, 2012).
3.2. Measuring CEO SNH
To measure the heterogeneity of a CEO’s social network, we gather personal information on
individuals in the CEO’s network from BoardEx. Four categories of heterogeneity are calculated
as follows:
A. Demographic heterogeneity: it is the average of two components—gender heterogeneity and
ethnicity heterogeneity. We measure gender heterogeneity by calculating an inverse
Herfindahl index (HHI) based on the percentage of female versus male in the network and
then taking the inverse. Specifically, HHI_gender=(female_ratio2 + male_ratio2).
Het_gender =1/ HHI_gender. The smaller the Het_gender, the more gender-diverse is one’s
social network. We measure the ethnic heterogeneity of a CEO’s social network by using
inverse Herfindal index based on the percentage of people from different international
regions. Specifically, HHI_ethnicity =(North_America_ratio2 + Latin_America_ratio2 +
Europe_ratio2 + Asia_ratio2
+ Africa_ratio2). Het_ethnicity=1/HHI_ethnicity. Finally,
demographic heterogeneity is measured by the mean of gender heterogeneity and ethnicity
heterogeneity. That is, Het_demographic = (Het_gender + Het_ethnicity )/2.
B. Intellectual heterogeneity: it is the average of three components—educational degree
heterogeneity, major heterogeneity, and school heterogeneity. Degree heterogeneity is
calculated by the inverse Herfindahl index based on the percentage of people with different
educational degrees. In particular, HHI_degree = PhD_ratio
2
+ Master_ratio
2
+
Bachelor_ratio2. Het_degree = 1/ HHI_degree. Major heterogeneity is calculated based on
the
percentage
of
people
(Business_Finance_ratio
2
with
different
majors.
2
Specifically,
2
+ Engineering_ratio + Liberal_Arts_ratio
HHI_major
=
2)
+ Law_ratio .
14
Het_major = 1/HHI_major. School heterogeneity is calculated based on percentage of people
graduated
from
the
same
school
as
the
CEO.
HHI_school
=
Same_school_ratio2+Different_school_ratio2. Het_school=1/HHI_school. Averaging three
indices, we get intellectual heterogeneity = (Het_degree + Het_major + Het_school)/3.
C. Profession heterogeneity: it is the average of two components—occupation heterogeneity and
industry heterogeneity. Occupation heterogeneity is calculated by the inverse Herfindahl
index based on the percentage of people with different occupations. We assume that different
job title reflects different expertise and management skills. In particular, HHI_occupation =
(CEO ratio)2 +(CFO ratio)2+ (Other executives ratio)2+(Board of directors ratio)2.
Het_occupation =1/HHI_occupation. Industry heterogeneity is calculated by the inverse
Herfindahl index based on percentage of people working in the same industry according to 2digit SIC code. HHI_industry = (same industry ratio)2+(different industry ratio)2.
Het_industry =1/HHI_industry. Averaging these two, we get Het_prof = ( Het_occupation +
Het_industry )/2.
D. International heterogeneity: this measure attempts to capture the diversity of CEO social
networks with international companies. We examine the headquarter locations of companies
at which the CEOs’ friends work. Based on the World Bank category of countries of income
level, we group headquarter countries into four groups: high income, upper middle income,
lower
middle
income,
and
lower
income
countries.
Using
Herfindahl
index,
HHI_international=(HighIncome_ratio2+UpperMidIncome_ratio2+LowerMidIncome_ratio2
+ LowIncome_ratio2). Het_international=1/HHI_international.
E. Overall heterogeneity: the average of above four heterogeneity indices.
Table 1 reports the summary statistics of CEO SNH by industries and firm types using HHI
measures (inverse of Heterogeneity). The reason we report HHI instead of heterogeneity is that
the HHI measure provides a better intuition to understanding the degree of diversity in a CEO’s
social network. For example, the HHI measure has a range of 0 to 1 and a value of 0 means that
everyone in the network has different backgrounds and hence the network is completely
15
heterogeneous. In contrast a value 1 means that everyone in the network has the same
background and hence the network is completely homogeneous.
In Table 1 we report CEO network HHI measures by firms’ industry, high tech type, whether
they have R&D investment, and whether they are multinational. On average, CEOs have lower
network HHI for professional networks and international networks. Since HHI is a reverse
measure of heterogeneity, this means that these two types of networks have higher diversity than
other types. In Panel A of Table 1, we report CEO network HHI by firm industries. We include
all industries except for finance (SIC1=6) and utility industry (SIC1=4). Results suggest that
network diversity does not have significant differences across industries. Panel B shows CEO
network HHI for high tech versus non-high tech firms. Following the literature, we define “high
tech” as 1 if the firm belongs to high tech and pharmaceutical industry (as classified by SIC2=48,
SIC2=73, SIC3=283) (Bodnaruk et al., 2013). We find that CEOs of high tech firms have lower
network HHI than non-high tech firms in terms of all types of networks. This indicates that high
tech CEOs tend to have more heterogeneous social networks than non-high tech firms. Panel C
examines CEO network HHI by firm R&D investment. We define “R&D” as 1 if a firm has nonzero R&D expenditures and 0 otherwise. Results suggest that firms with R&D investment are
associated with lower network HHI, which means higher CEO network heterogeneity. Finally,
Panel D reports CEO network HHI by foreign business. “Multinational” is defined as 1 if a firm
has foreign revenue in the testing year and 0 otherwise. As the results indicate, it is apparent that
multinational firms are associated with lower CEO network HHI for all types of networks. We
will test the determinants of CEO SNH in the regression analysis. For now, the statistics in Table
1 suggest that CEOs of high tech firms, R&D intensive firms, and multinational firms tend to
have more diversified networks than other firms.
< TABLE 1: Summary statistics of HHI measure of CEO SNH >
3.3. Other variables
16
A. Firm and CEO characteristics: we match BoardEx with Compustat using ticker and CIK
number to obtain firm characteristics. 1 Financial institutions, the utility industry, and
nonprofit organizations are excluded from the sample. Firm value is measured by Tobin’s Q.2
Our analysis also requires many other firm characteristics as specified in Table 2. We use
two measures of board heterogeneity. One is the percentage of female board members over
total number of board members. The second one is the percentage of minority board
members over the total number of board members. These variables are constructed using
board characteristics data from BoardEx. We obtain the following CEO characteristics from
BoardEx: age, gender, nationality, chairman position, and tenure. We construct a minority
indicator, which is 1 if the CEO’s nationality is U.S., 0 otherwise. For those CEOs with
missing information on the nationality, we create an indicator called minority missing. The
detailed variable definitions can be found in Table 2.
B. CEO turnover: we extract CEO turnover data from ExecuComp and then match with
BoardEx to obtain social network information of the CEOs. We only keep turnover cases,
where social network heterogeneity information is available for both departing CEOs and
new CEOs, thus allowing us to calculate differences in social network heterogeneity. To
measure the market reaction to the appointment of a new CEO, we examine cumulative
abnormal returns over three windows of [-1,1], [-2,2], and [-5,5]. Abnormal returns are
measured as the firms’ stock returns minus expected returns, which is estimated from a
market model with parameters estimated using daily stock returns over a period of [-255, 20] and using the CRSP value-weighted index to measure market returns. Following the CEO
turnover literature, we gather information regarding CEOs’ past employment history and
construct an indicator of whether the CEO is hired from outside the firm or not. Our
subsample on CEO turnovers consists of 114 turnover events between 2000 and 2010.
1
The matching needs to be done with care. In some cases, the same firm can be assigned with different Board IDs as
its name can be referenced with slight variations in different sources (Ishii and Xuan, 2010). We therefore drop all
firms with the same CIK/ticker but different names to minimize the matching problem.
2
This is calculated as the book value of assets plus the market value of equity (price times number of shares
outstanding) less the sum of the book value of common equity and balance sheet deferred taxes all over assets.
17
C. Innovation variables: To test the hypotheses relating innovation, we further match our
sample with NBER Patent database (Hall et al., 2001). Our measure for innovation is
logarithm of patents applied by a firm in a given year. Notice that patents are not granted
immediately after applying, and that there is generally a two to three year lag between
applying and granting patents, sometimes even more. To avoid the truncation problems that
may arise from a lag between applying and granting patents, our sampling criteria for patent
count is during application year and not when patents are granted (Francis et al., 2012). We
identify 525 firms in our sample that have patent applications recorded in NBER Patent
database. In addition to these firms, we also include firms that have zero patent applications
but also have non-zero R&D expenditure. Overall, our subsample for the innovation channel
test has 1759 firms.
D. M&A characteristics and performance measures: For the analysis on M&A investment, we
extract merger and acquisition deal data from Securities Data Corporation (SDC) and match
with our main sample. Deals are selected based on the following criteria: (1) announcement
dates are between 2000 and 2010; (2) transaction form is categorized as “Merger,”
“Acquisition,” “Acquisition of Assets,” or “Acquisition of Majority Interest”; (3) the deal is
complete; (4) the acquirer’s initial stake in the target firm is less than 50%, and the final stake
is more than 50% after merger and acquisition; (5) deal value is more than $1 million; (5)
acquirers have no multiple mergers in the same year; (6) acquirers have stock return data
available in CRSP. Finally, our M&A subsample contain 1300 deals. To measure M&A
performance, we examine market performance in both the short-run and the long-run. Shortrun performance is measured by the 5-day cumulative abnormal return (CAR) during the
event window of [-2, 2] experienced by the acquiring firms. Event day 0 is the acquisition
announcement date provided by SDC. Expected returns are estimated from a standard market
model over the period from [-210, -11] with the CRSP value-weighted return as the market
return. Long-run performance is measured by buy and hold abnormal returns (BHAR) over
the 3 year period after M&A announcement. In particular, BHAR=Π(1+Ri)-Π(1+Rm), where
18
Ri is monthly stock returns of the acquirers and Rm is the monthly value-weighted market
return.
<TABLE 2: Variable definitions >
3.4. Sample descriptive statistics
Table 3 reports the summary statistics of all other variables used in our analysis. The
main sample consists of 3506 CEOs from 2484 U.S. public firms over the 2000-2010 period.
Panel A reports the measures for demographic, intellectual, professional, and international
heterogeneity. Two other social network measures are network size and network centrality.
Network size is small, which reflects the sample screening limitation. We require that the person
connected with the CEO be at a senior position at other firms to maximize the probability that
two individuals actually have had informative conversations (Fracassi, 2012). Hence, the results
will not overestimate the effect of social networks. Panel B reports summary statistics for firm
characteristics in addition to board characteristics. We describe the variable definitions in Table
2. The average firm value is relatively high, with Tobin’s Q of 1.961. The average firm size is
about 2665 million spanning from 40.3 million for the 5 percentile to 13 billion for the 95
percentile. The large range of firm size in our sample is due to the fact that BoardEx provides
CEO information not only for S&P1500, but also broad cross-sections of firms of smaller sizes.
As for board characteristics of our sample firms, on average there are 9% of board members are
females and 6% have minority directors. The patent measure is number of patents applied for by
a firm in a given year. We assume that firms with positive R&D investment but no recorded
patent in the NBER database have zero patents. The mean patent application number in our
sample is 26.3.
Panel C focuses on the characteristics of CEOs. The average CEO age is 55.5 and tenure
is 10.5 years. Among our sample CEOs, 3% of CEOs are female CEOs and 0.6% are minority
CEOs. 32.2% of CEOs do not have nationality information and we treat them as non-minorities,
and we create a dummy variable of minority missing in the regression as control. . On average,
54.8% of CEOs are also chairman of the board. To study the CEO appointment announcement
19
effects, we identify 114 cases of CEO turnover in which we have social network information for
both the new incoming CEO and the previous CEO. Among the new incoming CEOs, 60% are
hired from outside of the firm. To measure market reactions to the new CEO appointment
announcement, we obtain cumulative abnormal returns over three windows of [-1,1], [-2,2], and
[-5,5].
Panel E reports descriptive statistics of variables used in the analysis for M&A
investment. We describe the variable definitions in Table 2. Our subsample for M&A analysis
contains 1370 deals. The average of acquirers’ 5-day CAR [-2, 2] is 0.2%, which is consistent
with prior studies (e.g., Masulis et al., 2007). The average of long-run performance, as measured
by one year BHAR, is about -7.2%. We measure the return-based run-up as the CRA of the
acquirer’s stock for trading days [-210, -11] relative to the M&A announcement date. The
average run-up is -0.7%. Moreover, in our M&A sample there are 28.9% diversified deals, 5.4%
tender offers, 11% all stock deals, and 29.6% mixed payment deals. As for target type, 48.7%
target firms are private firms, 23.9% are publicly traded firms, and the rest are subsidiaries.
<TABLE 3: Summary statistics >
4. Empirical Results
4.1. CEO SNH and firm value: baseline regressions
We begin estimating the effects of CEO social network heterogeneity on firm value using OLS
regressions. The dependent variable is firm value, as measured by Tobin’s Q, and independent
variables of interests include five CEO SNH indices. We control for firm characteristics
including size, leverage, capital expenditure, cash flow, R&D, and the diversity of board of
directors. In addition, we also control for network size and centrality to tease out their potential
effects on firm value. CEO SHN is measured for demographic, intellectual, professional,
international, and overall heterogeneity. Year dummies and industry indicators at 1-digit SIC
code are also included (but not written in the equations due to space constraints). Robust
standard errors are clustered at firm level. The estimation equation is as follows:
20
Q i,t = α0 + α • (CEO SNH)i,t + b • (Network measures)+  • (Firm and CEO char.)i, t +i,t
Table 5 reports the baseline OLS results. We find that intellectual, profession, and
international diversity are significantly and positively related to Tobin’s Q. However, the effect
of demographic diversity on firm value is not statistically significant. These results suggest that
CEOs with more diverse social connections create more firm value compared with CEOs with
less diverse social ties. And the benefits of diversity mainly come from knowing people with
different education backgrounds, diverse professional skills set, and multiple international
exposures. The role of demographic diversity does not seem to be as significant as other aspects
of diversity. In the last column, we regress overall heterogeneity index on firm value. Overall
index is the average of the four individual heterogeneity indices. The finding suggests that
average heterogeneity index is strongly associated with higher firm value. The coefficient is
0.532, which is statistically significant at 1% level. It indicates that if a CEO could increase his
or her social network heterogeneity by 10%, Tobin’s Q could increase by 4%.3 For an average
firm of our sample with total assets of 2665 million, 4% increase in Q indicates about 106
million increases in market value. This effect is strongly significant in the economic sense.
Using the same calculation approach, we report the economic interpretations other coefficients in
Table 6.
Among these diversity measures, international exposure has the largest impact on firm
value ($96 million increase in market value following a 10% increase in international exposure),
followed by professional heterogeneity ($30 million increase in market value following a 10%
increase in professional diversity) and intellectual heterogeneity ($28 million increase following
a 10% increase in intellectual diversity). Besides the demographic diversity, all the other
diversity measures play a nontrivial economic impact on firm value. We also compute the
economic impact of social network centrality on firm value using the coefficient from Column
(5). It shows that as centrality increases by 10%, firm value increases by 16 million for an
average firm in the sample. Compared with this magnitude, overall heterogeneity has much
3
We calculate this percentage change as: 10% * Mean (Het-overall) * β (Het-overall) / Mean (Q) =
0.1*1.467*0.532/1.961=4%
21
bigger economic impacts on firm value. These results highlight our argument that although
being in a central network position is important, knowing different people with diverse education
backgrounds and working experiences can be more crucial to create firm value.
< TABE 5: CEO SNH and firm value: baseline results>
< TABE 6: Economic interpretations of results in Table 5>
4.2. CEO SNH and firm value: instrumental variables approach
The relationships between CEO SNH and firm value can be spurious due to the possibility that
(1) better performing firms can select CEOs who have strong social connections (reverse
causality), and (2) certain firm characteristics can simultaneously affect CEOs’ choices of social
network and firm value (simultaneity bias). This means that firm value and CEO social network
can be determined in equilibrium so that it is difficult to interpret which is causing which. In our
case, the endogeneity problem, if not corrected, could cause our results be overestimated,
meaning that we are more likely to find CEO SNH to be positively associated with higher
Tobin’s Q.
An ideal way to deal with the endogeneity issue in our case is to find an exogenous shock
to CEO social network, which is plausibly unrelated to firm performance. Our main
identification strategy is to look at CEOs’ network contacts that have either died or retired during
the testing year of firm value. Based on Fracassi and Tate (2012), death of network ties provide
an ideal shock to the network, as it is less likely to be anticipated and is unrelated to firm
performance. Retired ties, however, may be anticipated, and are less likely to be replaced
immediately. Hence, retired ties can also be considered as an exogenous shock that changes
network ties directly but not firm value directly. To construct the instrument, we first count—for
each CEO each year—the number of deaths and retirements of individuals who belong to the
CEO’s social network. Additionally, to be taken into consideration, we also require that the
deceased and retired individuals have different backgrounds in at least one of the categories of
our heterogeneity measures (e.g., demographic, intellectual, profession, or international
experience). This is to ensure that death or retirement is a valid shock not only to the network
22
ties but also to the heterogeneity of ties. Finally, we divide the number of qualified deaths or
retirements by total number of social ties of the CEO to obtain a standardized measure, as a
larger network can give rise to more death and retirement. The model specification of 2-SLS IV
estimations is as follows:
CEO SNH =α0 + α• (Deceased or retired network ties)i,t+ •(Firm and CEO char)i, t+i,t(2-1)
Q i,t = β 0 + β • (Predicted SNH )i,t +  • (Firm and CEO char.)i, t-1 +i,t
(2-2)
Table 7 reports our results of 2-SLS IV estimations. 1st stage regressions results are
reported in Columns (1) (3) (5) (7) (9) for different types of CEO SHN, respectively. We find
that the level of CEO SNH significantly reduces following the death and retirement of network
connections that have different attributes than the CEO. To test if the instrumental variable has a
good predictability of CEO SNH, we report the F-statistics of the 1st stage regressions. Fstatistics across all 1st stage regressions are greater than the cut-off value of 10, which suggests
that our instrument is relevant and does not suffer from weak instrument concern (Staiger and
Stock, 1997). The results of the second stage regressions show that
CEO SNH for demographic, intellectual, professional, and international exposure are all
positively and significantly related to firm value. Comparing with the baseline results,
demographic diversity appears to have a significant effect on firm value in the IV estimation.
This means that after correcting for the endogeneity issue, demographic diversity of CEO social
ties can increase firm value. The overall heterogeneity measure, in both baseline and IV
estimation show a strong and significant coefficient with Tobin’s Q. Hence, we conclude that
having a CEO who has more heterogeneous social connections is value-added to shareholder
wealth.
<Table 7: Effect of CEO network heterogeneity on firm value: IV approach>
4.3. Event study on the market reaction to CEO appointment
We further address the endogeneity problem more generally by examining the market reaction to
the new CEO appointment. In particular, we compute the social network heterogeneity for the
new CEO and compared the score with the previous CEO. Two groups of firms are then formed:
the first group includes incoming CEOs with higher heterogeneity scores than their predecessors
23
(the heterogeneity-increasing group); the second group includes incoming CEOs with the same
or lower heterogeneity scores than their predecessors (the heterogeneity-decreasing group). Our
aim in this section is to investigate whether market reactions to CEO appointment can be
explained by the social network heterogeneity differences between the new CEO and the old
CEO. If our previous findings are robust, we expect that the market reaction will be higher when
the new CEO has greater social network heterogeneity.
While reverse causality is not a concern in the event study framework (as it is hard to
argue that higher market reaction leads to more heterogeneous CEO social networks), another
issue cannot be ignored:
the omitted variable problem. That is, some underlying firm
characteristics, such as size, capital structure, capital expenditure, cash flow and R&D
investment can affect both CEO social network and market reaction to the new CEO
appointment. In addition, personal characteristics of the new CEOs, including age, tenure,
chairman position, can explain market reactions, as well as social networks. To correct for any
endogenous selection on observables, we adopt the propensity score matching techniques to
match firms with similar firm characteristics and CEO characteristics based on the following
parameters:
logarithm of total assets, leverage, capital expenditure, cash flow, and R&D
intensity. According to the previous literature, whether the new CEO is promoted from inside of
the firm isa key factor in influencing the market’s perception of future firm performance. To
remove the potential market reaction differences that are driven by these CEO characteristics, we
also include these two CEO characteristics as matching criteria. As Panel A of Table 8 reports,
we compare the mean difference of the matched pairs and find that matched firms have no
differences in selected firm characteristics, and that the newly hired CEOs are also similar to
each other in terms of key experience variables. The only difference between the matched firms
is that one firm belongs to the heterogeneity-increasing group, while the other belongs to the
heterogeneity-decreasing group.
Panel B of Table 8 reports the results of the event study on CEO appointment. We report
cumulative abnormal returns (CAR) over the 3-day window [-1, +1], where day 0 is the date at
which the firm announces the new CEO appointment. As an alternative, we also compute a 5-day
24
window of [-2,2] and a 10-day window of [-5,5]. Our results for different event windows
consistently show that firms in which the old CEO is replaced with a new CEO with greater
social network heterogeneity experience a positive market reaction, which is statistically
significant at 10% level for the [-2, 2] event window, and become more significant for the [-5,
5]event window. In contrast, the firms where the new CEO has less heterogeneous social
networks turn out to have insignificant market reactions. Comparing the two groups, the
heterogeneity-increasing group has higher CAR than the heterogeneity-decreasing group. The
differences of CAR between the two groups are statistically significant at 10% level for 3-day
and 10-day event windows and 5% level for the 10-day event window. Thus, CEOs with more
heterogeneous social ties appear to increase shareholder wealth.
<Table 8: Investor response to CEO appointment announcement>
4.4. Exploring the channel through innovation
In order to better understand how CEO SNH enhances firm value, we further explore potential
channels. We first investigate whether heterogeneity of social network enhances innovation.
Social network theorists have documented the role of social networks in knowledge diffusion
(Goyal and Moraga-Gonzales, 2001). Innovation literature emphasizes knowledge heterogeneity
as a wellspring for creativity and innovation (Hargadon and Sutton, 1997; Galunic and Rodan,
1998). In addition, Hall et al. (2005) demonstrate a positive link between innovation and firm
value. In particular, in using patent as a proxy for knowledge assets, the paper shows that market
valuation is higher for firms that have more patents. In exploring the determinants of innovation,
researchers further show that an environment that promotes inter-firm knowledge learning and
technology sharing is crucial for innovation. For example, Gomes-Gasseres et al. (2006)
document that alliance partners tend to have greater knowledge flow and innovation. Rodan and
Gulunic (2004) argue that managers’ access to diverse knowledge is equally important for
innovation performance. Taking the insights offered by these papers, we hypothesize that CEO
SNH leads to more firm innovation, and that acting on innovation increases firm value. To test
this hypothesis, we run simultaneous equations on CEO social network heterogeneity,
innovation, and firm value. The model is specified as (3-1)-(3-3). In the first stage we use death
25
and retirement of network ties as instruments and receive a predicted value of network
heterogeneity. The reason to do so is that there could be endogeneity problems associated with
CEO social network and innovation (e.g., reverse causality). That is, more innovative firms tend
to hire CEOs who have heterogeneous social ties. In the second stage, we run simultaneous
equations model for (3-2) and (3-3) using seemingly unrelated regressions.
CEO SNHi,t =α0 + α• (Deceased or retired network ties)i,t+ •(Firm char)i, t+i,t (3-1)
Innovation i,t =α0 +α• (Predicted SNH)i,t +  • (Firm char)i, t + i,t
(3-2)
Q i,t = β 0+ β •(Innovation)i,t-1+ α•(Predicted SNH)i,t+ •(Firm char.)i t +i,t
(3-3)
Columns (1) and (2) in Table 9 report the results on innovation channel. We find that
CEO social network heterogeneity has a positive impact on innovation, as measured by
logarithm of patents. And innovation has a positive impact on firm value, as indicated by the
results in Column (2). Moreover, in Column (2) we also find that the effect of CEO social
network heterogeneity on firm value becomes less significant. This means that CEO social
network heterogeneity has a direct impact on firm value through the channel of its impact on
innovation. Once we run the simultaneous equation model to control for the channel of
innovation, CEO social network heterogeneity no longer significantly affects firm value. In
Columns (3) and (4) we repeat the same analysis for intellectual heterogeneity, and in Columns
(5) and (6) we repeat the analysis for professional heterogeneity. Interestingly, we find that these
two types of heterogeneity play a significant role in enhancing innovation and firm value through
innovation.
<Table 9: Regression results relating innovation channel >
4.6. Exploring the channel through foreign sale growth
Next, we investigate whether CEO SNH contributes to firm value through its impact on foreign
business generation. Entering a foreign market is a process that compounds the complexity of all
managerial tasks, especially cultural know-how (Prahalad, 1990; Carpenter and Sanders, 1998).
Since social networks serve as an important medium for managers to exchange knowledge and
experience, we believe that CEOs who have exposure to diverse groups of people are likely to
26
know more about foreign market information, reach a network of foreign contacts, identify good
opportunities, and push sales. We test this hypothesis by examining the impact of CEO SNH on
foreign sale growth. Similar to the analysis of innovation channel, the model for analyzing
foreign sale channel is specified as follows:
CEO SNHi,t =α0 + α• (Deceased or retired network ties)i,t+ •(Firm char)i, t+i,t (4-1)
Foreign sale growth i,t =α0 + α• (Predicted SNH)i,t +  • (Firm char)i, t + i,t
(4-2)
Qi,t =β 0+ β•(Foreign sale growth)i,t-1+α•(Predicted SNH)i,t+ •(Firm char.)it,+i,t(4-3)
Columns (3) and (4) of Table 9 reports the regression results for (4-1) – (4-3). Our
findings show that CEO social network heterogeneity has a positive and significant impact on
foreign sale growth. Once we control for foreign sale growth channel in the regression, CEO
social network heterogeneity loses significance on its own. This result suggests that CEOs with
heterogeneous social ties promote foreign sales, and by acting on foreign sales, firm value is
increased. We repeat the same analysis for demographic heterogeneity and international
heterogeneity in Columns (3) and (4) for demographic heterogeneity and Columns (5) and (6) for
international heterogeneity. The findings suggest that two types of heterogeneity play a
significant role in pushing foreign sale growth and enhancing firm value through foreign sale
channel.
<Table 10: Regression results relating foreign sale channel >
4.7. Exploring the channel through investment efficiency
Finally, we test whether CEO SNH affects firm value through its impact on corporate
investment. We argue that a heterogeneous social network is beneficial for a CEO to make better
decisions because such diversity is likely to provide the CEO with a wider breadth of
information sources and skill sets. Moreover, a heterogeneous social network could simulate
debate about the appropriateness of a business strategy, which allows the CEO to gain multiple
perspectives and alternative solutions (Wiersema and Bantel, 1992; Watson et al., 1993). Hence,
we expect that CEOs with greater diversity of networks are more likely to choose positive NPV
projects and make better investment that add value for the firms.
27
We start our analysis focusing on investment efficiency. We follow the literature to
measure investment efficiency. Researchers have used the change in investment sensitivity to Q
as a measure of investment efficiency. A positive change in the Q-sensitivity is interpreted as an
increase in investment efficiency. For example, Gertner et al. (2002) examine the investment
behavior of firms before and after spin-off from their parent companies. Dittmar and Shivdasani
(2003) investigate corporate divestiture and its impact on efficiency of segment investment.
Xuan (2009) examines how specialist CEOs affect internal capital allocation within a multisegment firm. His empirical approach is to first estimate investment efficiency using investmentQ sensitivity, and then to interact Q with specialist CEO. The coefficient of the interaction term
captures the changes of investment efficiency conditional on CEO type. Chen et al. (2007)
examine whether information content of stock prices and its impact on investment price
sensitivity. Using the same approach of interacting Q with information content, they demonstrate
that stocks with higher information content could provide managers more information and
improve their investment decisions.
Our empirical models follow the above literature. In the first step, we estimate
investment-Q sensitivity based on the typical Fazzari et al. (1988) investment equation as shown
in equation (5),
Ii,t/TAi,t-1= β0 + β1Qi,t-1 + β2(CFi,t /TAi,t-1)+ + •(Firm char)i, t +
y (Firm and year fixed effects)i, t +i,t
(5),
where investment (Ii,t) scaled by the lagged book value of assets (TAi,t-1) is regressed on lagged
Tobin’s Q (Qi,t-1 ) and cash flows (CFi,t) scaled by the lagged book value of assets (TAi,t-1).
When we calculate our measures of investment, we consider overall capital expenditure, which
includes capital expenditures, R&D expenditures, and acquisition expenditures. Moreover, we
also investigate acquisition expenditure as a separate measure because merger and acquisition is
often the largest capital expenditure in a firm if it happens. Both investment measures are scaled
by lagged total assets. Firm fixed effects and year fixed effects are included in order to control
for omitted variables over time that affect the investment level. The coefficient β1 captures the
investment-Q sensitivity. However, β1 is not the focus of our paper. Our goal is to examine
28
whether the level of investment efficiency (measured by the investment-Q sensitivity) depends
on CEO social network heterogeneity (CEO SNH).
To do so, we adopt a second step, in which we interact CEO SNH with Tobin’s Q. The
coefficient of CEO SNH*Q
t-1
will capture the effect of CEO network heterogeneity on the
sensitivity of investment to Q. This approach is similar to Xuan (2009), who examines the effect
of specialist CEOs on internal capital allocation efficiency, and Chen et al. (2007), who analyze
the effect of private information private information in price on the sensitivity of investment to
price. The model specifications are expressed in equations (6-1) and (6-2). In equation (6-1), we
use IV approach to the exogenous measure of overall CEO SHN. Using the predicted value of
CEO SNH in equation (6-2), we examine the interaction between CEO SNH and Q. And the
coefficient β3 is our main focus.
CEO SNHi,t =α0+ α• (Deceased or retired network ties)i,t+ •(Firm char)i, t+i,t (6-1)
Ii,t / TAi,t-1 = β0 + β1 Qi,t-1 + β2 (CFi,t / TAi,t-1)+ β3Qi,t-1 * CEO SNH + β4CEO SNH +
y (Firm and year fixed effects) + •(Firm char)i, t +i,t
(6-2)
The results are reported in Table 11. Column (1) reports the basic investment equation for
the overall capital expenditure. Tobin (1969) shows that Q, a proxy for investment opportunity,
is a predictor of investment. Hence, there should be a positive relationship between the stock
prices and the level of investment, which means that the β1 coefficient in equation (5) should be
positive and significant. We find a consistent result that Qi,t-1 is positively associated with total
capital expenditure ratio, with the coefficient for Qi,t-1 is 0.021signifciant at 1% level. To further
test our hypothesis on CEO SNH, we focus on Column (2), which includes additional interaction
term Het-overall-hat* Tobin’s Qi,t-1. Note that Het-overall-hat is the predicted value of CEO
SNH from first stage of IV estimation using percentage of deceased or retired social ties as an
instrument (equation 6-1). The coefficient of the interaction term is estimated at 0.299 with tstatistic of 2.739, which is statistically significant at 1% level. This suggests that the investmentQ sensitivity is higher when CEOs have more heterogeneous social connections. This supports
our hypothesis. Based on the prior literature, our regressions include the following set of control
variables: Inverse total asset, measured by 1/logarithm of total assets, leverage, and cash flow.
29
We use inverse total asset based on the argument by Chen et al. (2007) that both the dependent
variable (investment) and the regressor Q are scaled by last year’s book assets. Therefore, using
reverse asset can isolate the correlation between investment and Q induced by the common
scaling variable. For all regressions, firm year fixed effects and firm fixed effects are included to
control for omitted time invariant firm characteristics and economic changes over time.
Columns (3) and (4) repeat the same analysis but focus on acquisition expenditure ratio
instead of total capital expenditure. The purpose is to examine whether CEO SNH plays a
stronger/weaker role in the efficiency of M&A investment. We can see that in Column (3), the
coefficient of Qi,t-1 on acquisition investment ratio is positive, meaning that growth opportunity
increases acquisition related investment. This is consistent with the general view that investment
opportunity is a predictor of investment (Tobin, 1969). In Column (4), we find that the
interaction term Het-overall-hat* Tobin’s Qi,t-1 has a coefficient of 1.043, which is strongly
significant at 1% level. This suggests that the acquisition investment efficiency improves
significantly when CEOs have more heterogeneous social connections. Again, this supports our
hypothesis,
<Table 11: CEO SNH, investment efficiency, and firm value >
4.8. Exploring the channel through M&A
Acquisitions are one of the largest forms of corporate investment. These investments also tend to
be inefficient due to the inherent agency problems between managers and shareholders in large
public operations (Jensen and Meckling, 1976). For example, it is well documented that
managers prefer M&A investment not because they are positive NPV projects, but because they
bring significant personal benefits from empire building (Jensen, 1986; Lang et al., 1991). And
in this case, M&A is inefficient investment and is value destroying to shareholder wealth.
Morck et al. (1990) identify that diversifying M&A is one of the types of acquisition that
especially hurts shareholder wealth.
Our hypothesis regarding CEO SNH and investment argues that access to a
heterogeneous social network provides CEOs with a wider range of information and resources,
which is crucial for them to identify good investment opportunities. As we have found in the
30
previous test, CEO SNH is positively associated with acquisition related to investment
efficiency. In the following test, we go a step further to examine how CEO SNH affects value
effect of M&A. To measure wealth effect, we compute both short-run market reaction to the
M&A announcement and long-run post-merger performance. Specifically, we measure acquirer
by computing 5-day cumulative abnormal return in the event window of [-2, 2]. We estimate
expected returns from a standard market model over the period from [-210, -11] with the CRSP
value-weighted return as the market return. As Panel E of Table 3 shows, the average 5-day CAR
in the M&A sample is 0.2%. The magnitude is comparable with Masulis et al. (2007). Long-run
performance is measured by buy and hold abnormal returns (BHAR) over the 3 year period after
M&A announcement. In particular, BHAR=Π(1+Ri)-Π(1+Rm), where Ri is monthly stock returns
of the acquirers and Rm is the monthly value-weighted market return. After a series sample
selection and matching procedure, our M&A sample includes 1300 deals from 2000 to 2010.
In the regression analysis, we relate M&A performance measures to acquiring firms’
CEO SNH. Table 12 reports the regressions results relating CEO SNH and M&A performance.
In Column (1) we show that CEO SNH is significantly and positively associated with abnormal
returns of the acquirer in the 5-day window around the announcement. Note that the CEO SNH
measure we use here is the predicted value from the IV regression using number of death and
retirement of connections as instrument. Hence, there is less concern of an endogeneity problem.
In Column (2) we include diversified M&A and its interaction with CEO SNH. The purpose is to
examine whether CEOs with more heterogeneous networks make better M&A investment
decisions, especially for diversified deal. Our hypothesis says that heterogeneous social
connections bring different perspectives and market information from different areas of expertise
and industries. Hence it might be especially beneficial for making the right decisions on
diversified M&A deals. Our findings as reported in Column (2) support this hypothesis. The
indicator of diversifying M&A has a significant and negative coefficient on abnormal returns.
However, the interaction between CEO SNH and diversifying M&A is significantly and
positively associated with CAR [-2, 2]. In Columns (3) and (4) we analyze the effect of CEO
SNH on long-run performance of M&A, measured by 3-year BHAR of acquirers. The findings
31
are consistent with short-run market reaction. In particular, Column (3) shows that CEO SNH
has a significant and positive relationship with BHAR. Column (4) further reports that the
beneficial role of CEO SNH on long run performance is more pronounced for diversifying deals.
In the regression, we include a wide array of acquirer- and deal-specific characteristics.
For acquirer characteristics, we control for firm, leverage, market-to-book ratio, profitability,
sale growth, and pre-merger stock price run-up as measured by cumulative stock returns over the
[-210, -11] window (Masulis et al., 2007). For these control variables, our estimates are
consistent with prior studies (e.g., Moeller et al., 2004; Masulis et al., 2007). In particular, we
observe that acquirer firm size has a significant and negative relationship with CAR. Acquirer
leverage has a positive and significant effect on CAR, which supports the argument that leverage
serves as an important governance mechanism as higher debt ratio reduces managers’ incentives
to make inefficient investment. We find that market to book ratio has a negative, albeit
insignificant, effect on bidder returns. We do not find pre-merger acquirer stock run-up to be
significantly associated with bidder returns. However, prior operating profitability has a
significant effect on bidder returns in the short run.
Deal characteristics that we control for include payment method, public status of target
firms, relative deal size, tender offer, and whether it is a diversified M&A, measured at 1-SIC
level. Prior literature has well documented that acquirers experience significantly negative
abnormal returns when they are buying publicly traded firms as opposed to private targets (Fuller
et al., 2002). As shown in Table 12, we find strong evidence supporting prior literature. The
public target variable is significantly and negatively associated with market reaction to the M&A
announcement. It is also negatively associated with long run performance of M&A. Consistent
with Moeller et al. (2004), our results also imply that subsidiary target—the omitted group,
because we create three groups based on target type: public, private, and subsidiary—should be
associated with the highest abnormal bidder returns both in the short run and the long run. Our
findings in Table 12 also support the existing evidence that payment method plays an important
role in affecting abnormal returns in the short run (Chang, 1998; Fuller et al., 2002). For
example, we find stock deal M&As are associated with significantly negative announcement
32
abnormal returns and that mixed payment method has no significant coefficient. And the omitted
group—cash payment—should be positively associated with abnormal returns. These are
consistent with the adverse selection problem in equity issuance analyzed by Myers and Majluf
(1984). We control for relative deal size, but only find it significantly and positively associated
with long run performance of M&A. This is consistent with Asquith et al. (1983) and Moeller et
al (2004). We also control for tender offer, which is a dummy variable equal to 1 if the deal is
categorized as tender offer in SDC. We do not find it to have a strong effect on abnormal return,
except for Column (2) when we control diversified M&A. Lastly, in Columns (2) and (4) we test
the differential effects of CEO SNH on diversified M&A deals versus focused M&A deals.
Diversified M&A is defined as 1 if the acquirer and target are in the same industry as measured
by 1-SIC code. We find strong evidence that diversified M&A is significantly associated with
negative abnormal returns both in the short run and in the long run. Overall, the coefficients on
the deal characteristics are consistent with existing literature.
<Table 12: Regressions on CEO SNH and M&A performance >
5. Conclusion
While existing literature has documented various benefits and costs of CEO social network, little
attention is paid to the composition and heterogeneity aspect of CEO social network. This is a bit
surprising, since who is in a CEO’s social network clearly has an impact on what information
and resources he or she can obtain. Joining a heterogeneous group of people could offer diverse
knowledge, new perspectives, and multiple problem-solving options that enrich the CEO’s
knowledge set and improve decision-making. In contrast, the benefit from a homogeneous social
network can be marginal.
We examine this issue by empirically testing the impacts of CEO social network
heterogeneity on innovation creation, new revenue generation from foreign markets, and
corporate investment decisions. Our study measures different aspects of heterogeneity such as
demographic, intellectual, profession, and international exposure of CEO social networks. We
find that CEO social network heterogeneity is positively associated with firm innovation, foreign
sale growth, investment efficiency, and M&A performance. Overall, CEO social network
33
heterogeneity significantly enhances firm value. We apply different approaches to deal with the
endogeneity problem and the results remain robust. These results overall are consistent with the
notion that greater heterogeneity allows for transfer of different knowledge, expertise, and
problem-solving skills between connected people and companies, which is value-added to the
firm. To the best of our knowledge, this paper is the first to manifest a positive link between
social network heterogeneity and firm value.
Our findings have a number of broad implications. The fields of economics and finance
have come to focus too much on quantitative skills and not enough on social capital. The
literature on CEO characteristics has also largely emphasized hard skills—such as education and
professional qualification—as key factors for managerial performance, but has overlooked
CEOs’ social network skills. Contemporary CEOs require a broader set of knowledge to
response to product innovation and increased competitive business pressure in the market. Yet
acquiring knowledge can be costly. Our findings suggest that a diverse social network provides a
CEO with exposure to different information and resources, which ultimately improves
managerial performance. Our results can hopefully encourage corporate shareholders to think
about how, given the changing face of the workforce and increasing competition from
international markets, social networks of upper management and board members can be valueadded for the company.
Moreover, policy makers are concerned about the increased diversity in the workplace.
Some claim that firms are pressured to hire minority workers due to ethical reasons rather than
profitability, while some argue that firms do not want to lose talented employees with varied
experience, knowledge, and cultural backgrounds because they can assist the firm in becoming
more successful in global marketplace. Through the lens of CEOs’ connections in the overall
labor market, our findings offer academic evidence that diversity and heterogeneity are tangible
assets that contribute to corporate profit. As pressure from global market competition and
product innovation increases, the heterogeneity of CEO social network will become more
important.
34
References:
Adams, R., Ferreira, D. 2009, Women in the boardroom and their impact on governance
and performance. Journal of Financial Economics, 94: 291-309.
Adams, R.B., Hermalin, B.E., and Weisbach, M.S., 2010. The role of boards of directors
in corporate governance: A conceptual framework and survey. Journal of Economic
Literature, 48(1): 58-107
Anderson, D. C., Reeb, D. B., Upadhyay, A., Zhao, W. 2011. The economics of director
heterogeneity. Financial Management, 40(1): 5-38.
Asch, S., 1951, Effects of Group Pressure upon the Modification and Distortion of
Judgment, in H. Guetzkow (ed.) Groups, Leadership, and Men. Pittsburgh: Carnegie
Press.
Ashraf, Q., Galor, O. 2013. The out of Africa hypothesis, human genetic diversity, and
comparative economic development. The American Economic Review, 103(1): 1-46.
Barnea, A., Guedj, I. 2007, CEO compensation and director networks, Working paper.
Barron, F. B., Harrington, D. M. 1981. Creativity, intelligence, and personality. Annual
Review of Psychology 32: 439–476.
Bennedsen, M., Perez-Gonzales, F., Wolfenzon, D. 2012. Evaluating the Impact of The
Boss: Evidence from CEO Hospitalization Events. Working Paper.
Bertrand, M., Schoar, A. 2003. Managing with Style: The Effect of Managers on Firm
Policies, Quarterly Journal of Economics 118(4), 1169-1208.
Bodnar, G. M., Weintrop, J. 1997. The valuation of the foreign income of US
multinational firms: a growth opportunities perspective. Journal of Accounting and
Economics, 24: 69-97.
Bodnaruk, A., Massa, M., Simonov, A. 2013. Alliances and corporate governance.
Journal of Financial Economics 107 (3): 671-693.
Burt, R.S. 1992. Structural Holes. New York: Cambridge University Press.
Butler, A.W., Gurun, U.G. 2012. Educational Networks, Mutual Fund Voting Patterns,
and CEO Compensation. Review of Financial Studies.
Cai, Y., Sevilir, M. 2012. Board Connections and M&A Transactions. Journal of
Financial Economics, 103: 327-349.
Calvo-Armengol, A., Jackson, M. 2004 .The Effects of Social Networks on Employment
and Inequality. American Economic Review, 94, 426.454.
Carter, D. A., Simkins, B. J., Simpson, W. G. 2003. Corporate Governance, Board
Diversity, and Firm Value. Financial Review, 38 (1): 33-53.
Carpenter, M. A., Sanders, WM. G.,1998, Internationalization and firm governance: The
roles of CEO compensation, top team composition, and board structure, Academy of
Management Journal, 41, 158:178.
35
Chen, Q., Goldstein, I., Jiang, W. 2007. Price informativeness and investment sensitivity
to stock price. Review of Financial Studies, 20(3): 619-650.
Cohen, L., Frazzini, A., Malloy, C. 2008. The small world of investing: Board
connections and mutual fund returns. Journal of Political Economy, 116(5): 951–979.
Cohen, L., Frazzini, A., Malloy, C., 2010. Sell side school ties, Journal of Finance 65(4),
1409-1437.
De Dreu, C. K., West, M. A. 2001. Minority dissent and team innovation: The
importance of participation in decision making. Journal of Applied Psychology, 86,
1191−1201.
Dittmar, A., Shivdasani, A. 2003. Divestitures and Divisional Investment Policies.
Journal of Finance, 58:2711–43.
Ellison, G., Fudenberg., D. 1993. Rules of thumbs for social learning. The Journal of
Political Economy 101(4), 612–643.
Ellison, G. Fudenberg., D. 1995. Word-of-mouth communication and social learning. The
Quarterly Journal of Economics 110 (1), 93–125.
Engelberg, J., Gao, P., Parsons, C. 2012. Friends with Money, Journal of Financial
Economics, 103(1): 169-188.
Engelberg, J., Gao, P., Parsons, C. 2013. The Value of a Rolodex: CEO Personal
Network and Compensation. Review of Financial Studies 26(1): 79-114.
Faleye, O., Kovacs, T., Venkateswaran, A. 2013. Do Better-Connected CEOs Innovate
More? Journal of Financial and Quantitative Analysis, forthcoming.
Fazzari, S., Hubbard, R. G., Petersen, B. C. 1988, Finance constraints and corporate
investment, Brookings Papers on Economic Activity 1, 141–195.
Fracassi, C. 2012. Corporate finance policies and social networks. AFA 2011 Denver
Meetings Paper.
Fracassi, C., Tate, G. 2012. External networking and internal firm governance. Journal of
Finance, 67 (1): 153-194.
Francis, B., Hasan, I., Wu, Q. 2014. Gender differences in financial reporting decisionmaking: Evidence from accounting conservatism. Contemporary Accounting
Research, forthcoming.
Francis, B., Hasan, I., Wu, Q. 2013. Professors in the Boardroom and their Impact on
Corporate Governance and Firm Performance. Working paper.
Fryer R, Loury G. 2013. Valuing Diversity. Journal of Political Economy. 121(4):747774
Galunic, D. C., Rodan, S. A. 1998. Resource recombinations in the firm: knowledge
structures and the potential for Schumpeterian innovation. Strategic Management
Journal 19(12): 1193–1201.
Goyal, S., Moraga-Gonzales, J. L. 2001. R&D Networks. RAND Journal of Economics,
32, 686-707.
36
Gomes-Casseres, B., Hagedoorn, J., Jaffe, A. B. 2006. Do alliances promote knowledge
flows? Journal of Financial Economics. 80(1): 5-33.
Granovetter, M. 1995. Getting a job: a study of contacts and careers, 2nd edition,
Harvard University Press, Cambridge.
Gul, F., Srinidhi, B., Ng, A. 2011. Does board gender diversity improve the
informativeness of stock prices? Journal of Accounting and Economics, 51 (3): 314338.
Hargadon, A., Sutton, R. I., 1997. Technology brokering and innovation in a product
development firm. Administrative Science Quarterly 42: 716–749.
Hall, B., Jaffe, A., Trajtenberg, M. 2001. The NBER patent citations data file: Lessons,
insights and methodological tools. Working Paper 8498, NBER.
Hall, B., Jaffe, A., Trajtenberg, M. 2005. Market value and patent citations. The Rand
Journal of Economics 36, 16-38.
Holzer, H., 1987. Hiring procedures in the firms: their economic determinants and
outcomes. In: M. Kleiner, ed., Human Resources and the Performance of the Firm,
Wis.: Industrial Relation Research Association, 243-274.
Horton, J., Serafeim, G. 2009. Security Analyst Networks, Performance and Career
Outcomes. Working paper available at SSRN: http://ssrn.com/abstract=1522313
Hong, H., Kubik, J. D., Stein, J. C. 2005. Thy Neighbor’s Portfolio: Word-of-Mouth
Effects in the Holdings and Trades of Money Managers. Journal of Finance, 60 (6),
2801-2824.
Hong, L., Page, S. E. 2001. Problem Solving by Heterogeneous Agents. Journal of
Economic Theory, 97 (1), 123-163.
Hwang, B., Kim, S. 2009. It pays to have friends. Journal of Financial Economics,
93(1):138–158.
Janis, I. L., 1982, Groupthink: Psychological Studies of Policy Decisions, 2nd ed.,
Houghton Mifflin Company.
Kaplan, S. N., Klebanov, M. M., Sorensen, M. 2012. Which CEO characteristics and
abilities matter?. The Journal of Finance, 67(3), 973-1007.
Liu, Y. 2010. The Impact of Networks on CEO Turnover, Appointment, and
Compensation. Available at SSRN: http://ssrn.com/abstract=1573244.
Masulis, R. W., Wang, C., Xie, F. 2012. Globalizing the boardroom—The effects of
foreign directors on corporate governance and firm performance. Journal of
Accounting and Economics, 53 (3): 527-554.
McPherson, M., Smith-Lovin, L. Cook, J. 2001. Birds of a Feather: Homophily in Social
Networks, Annual Review of Sociology 27, 415-44.
Pelled, L. 1996. Demographic diversity, conflict, and work group outcomes: An
intervening process theory. Organization Science, 7, 615−631.
37
Pfeffer, Jeffrey. 1983. “Organizational Demography.” pp. 299-357 in L.L. Cummings
and Barry M. Staw (eds.), Research in Organizational Behavior (Volume
5).Greenwich, CT: JAI Press.
Powell, W. W. 1990. Neither market nor hierarchy: Network forms of organization.
Prahalad, C. K., Hamel, G. 1990. The core competence of the corporation. Harvard
Business Review, 68(3): 79–90.
Reuber, A. R., Fischer, E. 1997. The influence of the management team’s international
experience on the internationalization behaviors of SMES. Journal of International
Business Studies, 28, 807-825.
Robinson, G., Dechant, K. 1997. Building a business case for diversity. Academy of
Management Executive, 11(3): 21-30.
Rodan, S., Galunic, C. 2004. More than network structure: How knowledge heterogeneity
influences managerial performance and innovativeness. Strategic Management
Journal, 25: 541-562.
Shalley, C.E., Gilson, L.L. 2004. What leaders need to know: A review of social and
contextual factors that can foster or hinder creativity. Leadership Quarterly, 15: 3353.
Shalley, C. E., Perry-Smith, J. E., 2008. The emergence of team creative cognition: The
role of diverse outside ties, socio-cognitive network centrality, and team evolution.
Strategic Entrepreneurship Journal, 1, 2, 23-41.
Shrestha, L. B. 2011. Changing Demographic Profile of the United States. DIANE
Publishing.
Staiger, D., Stock, J. 1997. Instrumental Varaiables Regression with Weak Instruments,
Econometrica, 557-586.
Tobin, J., 1969, A general equilibrium approach to monetary theory, Journal of Money,
Credit and Banking 1, 15–29.
Toossi, M. 2002. Century of Change: The US Labor Force, 1950-2050, A. Monthly Lab.
Rev., 125, 15.
Uzzi, B. 1996. The sources and consequences of embeddedness for the economic
performance of organizations: The network effect. American Sociological Review,
61(4): 674–698.
Uzzi, B. 1999. Embeddedness in the making of financial capital: How social relations and
networks benefit firms seeking financing. American Sociological Review, 64(4):
481–505.
Wagner, W. G., Pfeffer, J., O'Reilly III, C. A. 1984. Organizational demography and
turnover in top-management group. Administrative Science Quarterly, 74-92
Watson, W. E., Kumar, K., Michaelsen, L. K. 1993. Cultural diversity’s impact on
interaction process and performance: comparing homogeneous and diverse task
groups. Academy of Management Journal, 36:3, 590-602.
38
Watts, D. J. 2003. Six degrees - the science of a connected age. W. W. Norton &
Company.
Wiersema, M. F., Bantel. L. 1992. Top management demography and corporate strategic
change. Academy of Management Journal, 35:1, 91-121.
Williams, K. Y., O'Reilly, C. A. 1998. Demography and diversity in organizations: A
review of 40 years of research. Research in Organizational Behavior, 20, 77−140.
Xuan, Y. 2009. Empire-building or bridge-building? Evidence from new CEOs’ internal
capital allocation decisions. Review of Financial Studies, 22(12), 4919-4948.
39
Table 1
Descriptive statistics of CEO SHN by industry and firm type (HHI measures)
This table reports the descriptive statistics of the Herfindahl index (HHI) of CEO social networks. HHIdemographic measures how focused CEO social networks are in terms of their demographical attributes.
HHI-intellectual measures how focused CEO social networks are in terms of their education achievement.
HHI-profession measures how focused CEO social networks are in terms of their working experience.
HHI-international measures how focused CEO social networks are in terms of their international exposure.
Lastly, HHI-overall is average of all. Note that HHI are the inverse measures of heterogeneity (Het).
Theoretically, the minimum value of HHI can be 0, which means that the network is very heterogeneous
and everyone in the network has different attributes, while a maximum of HHI is 1, which means that the
network is very homogeneous and everyone has the same attributes. We report descriptive statistics of HHI
measures for the purpose of easy interpretation, but in the regression analysis we use Het, which is 1
divided by HHI. We report HHI measures by firms’ industry, high tech type, whether the firm has R&D
activities, and whether the firm is multinational. High tech is a indictor, which is equal to 1 if firms belong
to pharmaceutical industry and high tech industry as classified by SIC2=48, SIC2=73, SIC3==283. The
indicator of R&D is defined as 1 if a firm has non-zero R&D expenditure and 0 otherwise. Multinational is
defined as 1 if the firm has non-zero revenue from foreign countries and 0 otherwise.
Panel A: By Industry 1-digit SIC
HHIHHI-profession
intellectual
SIC1
HHIdemographic
HHIinternational
HHI-overall
0
0.880
0.825
0.695
0.669
0.704
1
0.935
0.761
0.611
0.675
0.697
2
0.898
0.732
0.611
0.666
0.682
3
0.916
0.757
0.622
0.670
0.691
5
0.923
0.777
0.659
0.670
0.704
7
0.919
0.761
0.614
0.672
0.684
8
0.909
0.754
0.639
0.671
0.702
High Tech
Panel B: By High tech (High tech=1 if SIC2=48, SIC2=73, SIC3==283)
HHIHHIHHIHHI-profession
demographic
intellectual
international
HHI-overall
0
0.916
0.756
0.629
0.670
0.694
1
0.908
0.753
0.604
0.668
0.678
R&D
0
1
Multinational
Panel C: By R&D (R&D=1 if R&D expenditure>0)
demographic
intellectual
international
HHI-overall
0.923
0.770
0.639
0.673
0.701
0.908
0.745
0.613
0.668
0.684
Panel D: By foreign business (Multinational=1 if foreign revenue>0)
demographic
intellectual
international
HHI-overall
0
0.927
0.782
0.645
0.676
0.703
1
0.905
0.734
0.607
0.666
0.682
40
Table 2: Variable Definitions
Variable
Definition
CEO social network characteristics
Data source
Het-demographic
Average index of gender and ethnicity heterogeneity of a CEO's
social network
BoardEx
Het-intellectual
Average index of education degree, major and school heterogeneity
of a CEO's social network
BoardEx
Het-profession
Average index of occupation and industry heterogeneity of a CEO's
social network
BoardEx
Het-international
Average index of heterogeneity with foreign companies from
different income and economic development group
BoardEx
Het-overall
Average of four heterogeneity indices above
BoardEx
Network size
Log of total number of social ties
BoardEx
Centrality
A measure of network position of the CEO, calculated by the degree
centrality of CEO's professional network
BoardEx
Tobin's Q
Firm characteristics
Total assets (#6) + market value of equity (#25*#199) - book value
of common equity (#60)- deferred taxes (#74) / Total assets (#6).
Compustat
Log(assets)
Log of total assets (#6)
Compustat
Leverage
Long term debt (#9) + debt in current liabilities (#) / total assets (#6)
Compustat
Capextoasset
Capital expenditure (#128) / total assets (#6)
Compustat
Cashflow
Operating income before depreciation (#13) / Lag of total assets
(#6).
Compustat
R&D intensity
R&D expenditure (#46) / total sales (#12)
Compustat
Total capital expenditure
ratio
Capital expenditure (#128) + R&D expenditure (#46) + acquisition
expenditure (#129) - cash receipts from sale of
property, plant, and equipment (# 107) / total assets of last year (#6)
Compustat
Acquisition expenditure
ratio
Acquisition expenditure (#129) /total assets of last year (#6)
Compustat
High tech
Indicator equal to 1 if SIC2 is 48 or 73, or SIC3 is 283
Compustat
Patent
Number of patents applied by a firm
NBER
Foreign sales growth
Percentage increase of foreign revenue (#273) from last year
Compustat
Board female ratio
Percentage of female board of directors
BoardEx
Board minority ratio
Percentage of minority board of directors
BoardEx
CEO personal characteristics
Age
CEO age
BoardEx
Female CEO
Indicator equal to 1 if a CEO is female, 0 otherwise
BoardEx
Minority
Indicator equal to 1 if a CEO's nationality is not US, 0 otherwise
BoardEx
Minority missing
Indicator equal to 1 if the information on CEO nationality is missing
BoardEx
Tenure
Indicator equal to 1 if a CEO has a doctor degree, 0 otherwise
BoardEx
41
Chairman
Indicator equal to 1 if a CEO is graduate from a Ivy League school,
0 otherwise
BoardEx
Outside hire
Indicator equal to 1 if the new CEO is hired from outside the firm, 0
otherwise
BoardEx
CEO turnover characteristics
Cumulative abnormal returns of [-1, 1], where 0 is the
announcement date of new CEO appointment
CRSP,
ExecuComp
CAR [-2,2]
CRSP,
ExecuComp
CAR [-5,5]
CRSP,
ExecuComp
CAR [-2,2]
Cumulative abnormal returns for the acquirer over [-2, 2], where 0
is the M&A announcement date
CAR [-1,1]
M&A characteristics
BHAR- 3year
Run-up
Diversify M&A
Tender off
Relative size
All stock payment
Mix stock and cash
payment
Private target
Public target
Relative size
Buy and hold abnormal returns of the acquirer over 3 year after the
M&A
Cumulative abnormal returns for the acquirer over [-210, -11],
where 0 is the M&A announcement date
Indicator equal to 1 if the acquirer is in different industry than
target, based on 1-digit SIC.
indicator equal to1 if the deal is recorded as a tender off in SDC, 0
otherwise
CRSP, SDC
CRSP, SDC
CRSP, SDC
SDC
SDC
Deal value over market value of the acquirer
SDC, CRSP
Indicator equal to 1 if the deal is 100% paid by stock, 0 otherwise
SDC
Indicator equal to 1 if the deal is paid by combination of cash and
stock, 0 otherwise
SDC
Indicator equal to 1 if the target firm is private, 0 otherwise
SDC
Indicator equal to 1 if the target firm is publicly traded, 0 otherwise
SDC
Number of M&A deals an acquirer has conducted in a year
SDC
42
Table 3: Summary Statistics
This table reports the summary statistics of CEO social network measures, firm and board characteristics, CEO
characteristics, and all other variables used in the analyses. Variable definitions are described in Table 1.
Variable
Het-demographic
Het-intellectual
Het-profession
Het-international
Het-overall
Network size
Centrality
Tobin's Q
Total assets (million)
Leverage
Capextoasset
Cashflow
R&D intensity
Board female ratio
High tech
Patent
Foreign sales growth
Total capital expenditure ratio
Acquisition expenditure ratio
Age
Female CEO
Minority
Minority missing
Chairman
Tenure
Outside hire
CAR [-1,1]
CAR [-2,2]
CAR [-5,5]
CAR [-2,2]
BHAR- 3year
Runup
Diversify M&A
Tender offer
Relative size
All stock payment
Stock and cash mix payment
Private target
Public target
N
Mean
S.D
P5
Panel A. CEO social network measures
10272
1.112
0.153
1.000
11336
1.394
0.292
1.000
10919
1.706
0.413
1.000
10248
1.505
0.138
1.000
8846
1.467
0.162
1.100
12366
2.061
1.082
0.693
12366
0.0004
0.0003
0.0000
Panel B. Firm characteristics
12366
1.961
1.170
0.604
12366
2665.4
5206.8
6.1
12366
0.197
0.185
0.000
12366
0.046
0.043
0.001
12366
0.134
0.153
-0.996
12366
0.285
3.291
12366
0.091
0.095
12366
0.062
0.135
6211
0.277
0.448
6211
26.256
105.829
8471
0.112
1.381
14046
0.147
0.140
8752
0.094
0.216
Panel C. CEO characteristics
9079
55.760
7.266
9079
0.031
0.173
12366
0.006
0.074
12366
0.322
0.467
12366
0.548
0.498
12366
10.497
6.819
114
0.596
0.493
Panel D. CEO announcement returns
114
0.004
0.060
114
0.004
0.069
114
0.016
0.094
Panel E. M&A sample
1370
0.002
0.078
1370
-0.072
0.616
1370
-0.007
0.169
1370
0.289
0.453
1370
0.054
0.226
1370
0.000
0.000
1370
0.110
0.313
1370
0.296
0.456
1370
0.487
0.500
1370
0.239
0.426
Median
P95
1.000
1.421
1.619
1.456
1.468
1.946
0.0003
2.000
2.419
2.978
2.300
2.130
5.875
0.0049
1.596
741.4
0.171
0.032
10.720
40877.0
0.985
0.305
0.154
0.518
0.000
0.000
0.000
0.000
0.000
-1.764
0.004
0.000
0.012
0.100
0.000
0.000
0.000
0.081
0.105
0.029
271.385
0.800
0.800
1.000
124.000
2.212
0.979
8.338
31.000
0.000
0.000
0.000
0.000
0.000
0.000
56.000
0.000
0.000
0.000
1.000
10.400
1.000
91.000
1.000
0.000
1.000
1.000
22.500
1.000
-0.081
-0.114
-0.111
0.004
0.002
-0.003
0.077
0.129
0.212
-0.404
-2.767
-0.858
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.001
-0.153
-0.012
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.488
3.599
0.809
1.000
1.000
0.003
1.000
1.000
1.000
1.000
43
Table 4: Correlation Matrix
This table reports correlation matrix among main variables.
1
1
1
2
Tobin's Q
2
Het-demographic
0.0116
1
3
Het-intellectual
0.0076
0.2065*
1
4
Het-profession
0.0408*
0.1661*
0.3812*
1
5
Het-international
0.0507*
0.3920*
0.3833*
0.1387*
1
6
Het-overall
0.0621*
0.4745*
0.7176*
0.7862*
0.5241*
1
7
Network size
-0.0096
0.4547*
0.5175*
0.4237*
0.3500*
0.4912*
1
8
Centrality
0.0043
0.1242*
0.3603*
0.2175*
0.1211*
0.2562*
0.4421*
9
Total assets
-0.0537*
0.1532*
0.2866*
0.1092*
0.1387*
0.2110*
0.3116*
10
-0.2119*
0.0092
0.0747*
0.0186*
0.0064
0.0117
0.0919*
11
Leverage
Capextoasset
0.0149*
-0.0321*
-0.0375*
-0.0168*
-0.0197*
-0.0315*
-0.0063
12
Cashflow
-0.0596*
0.0054
0.0280*
-0.0031
0.0328*
0.0037
0.0981*
13
R&D intensity
0.0864*
0.0409*
0.0032
0.0024
0.0097
0.0350*
-0.0136
14
Board female ratio
0.0098
0.1189*
0.1527*
0.0362*
0.1052*
0.1111*
0.1848*
15
0.0167*
0.1287*
0.1019*
0.0647*
0.0496*
0.1081*
0.1003*
16
CEO age
-0.1140*
-0.0317*
0.0525*
-0.0106
0.0005
0.0045
0.0733*
17
CEO female
0.0172*
0.0549*
0.0534*
0.0147
0.0357*
0.0598*
0.0504*
18
Minority
0.0353*
0.0228*
0.0041
0.0370*
0.0199*
0.0452*
-0.0087
19
Minority missing
-0.0206*
-0.0227*
-0.0692*
-0.0333*
-0.0403*
-0.0286*
-0.0837*
20
Chairman
-0.0530*
0.0443*
0.1071*
0.0250*
0.0501*
0.0523*
0.1520*
21
Tenure
0.1142*
0.0232*
-0.0596*
-0.0035
-0.01
-0.0179*
-0.0917*
10
11
12
13
14
Centrality
8
1
9
8
9
Total assets
0.4499*
1
0.1549*
1
11
Leverage
Capextoasset
0.1227*
0.0159*
0.0267*
0.0518*
1
12
Cashflow
0.1320*
0.1417*
0.0819*
0.2954*
1
13
R&D intensity
-0.0262*
-0.0354*
-0.0253*
-0.0276*
-0.2451*
1
14
Board female ratio
0.2094*
0.2394*
0.0650*
0.0344*
0.1206*
-0.0044
1
15
0.1274*
0.2221*
0.0167*
-0.0205*
0.0129
0.0064
0.0478*
16
CEO age
0.1152*
0.0714*
0.0910*
0.0168*
0.0926*
-0.0213*
-0.0067
17
CEO female
0.0633*
0.0252*
-0.0193*
0.0146
-0.0072
-0.0026
0.2878*
18
Minority
-0.0111
0.0103
-0.0272*
-0.0039
0.0221*
-0.0041
-0.0071
19
Minority missing
-0.1314*
-0.1562*
-0.0460*
-0.0484*
-0.0897*
0.0193*
-0.0873*
20
Chairman
0.1914*
0.1619*
0.1081*
0.0166*
0.1516*
-0.0319*
0.1011*
21
Tenure
-0.1355*
-0.0891*
-0.1038*
-0.0209*
-0.0986*
0.0195*
-0.0052
15
16
17
18
19
20
21
10
3
4
5
6
7
44
15
1
16
CEO age
0.0055
1
17
CEO female
-0.0086
-0.0567*
1
18
Minority
0.1897*
-0.0062
0.0442*
1
19
Minority missing
-0.0496*
-0.0664*
-0.0154*
-0.0516*
1
20
Chairman
0.0302*
0.2799*
-0.0485*
0.0135
-0.1148*
1
21
Tenure
-0.0156*
-0.9747*
0.0559*
0.0023
0.0649*
-0.2793*
1
45
Table 5: CEO SNH and Firm Value —OLS Regressions
This table reports the OLS regression results relating CEO social network heterogeneity and firm value.
The dependent variable is Tobin’s Q. Independent variables of main interest are five measures of CEO
social network heterogeneity, including demographic, intellectual, profession, international, and overall
heterogeneity. Control variables include CEO social network size and centrality, key firm financial
variables, board diversity, and CEO characteristics. Firm financial variables are measured at the
previous year before the testing year. Detailed variable definitions are available in Table 2. Year
dummies and industry indicators at 1-digit SIC code are included. Robust standard errors are clustered
by firm. Numbers in the parentheses are robust t-statistics. *, **, and *** denote significance at 10%,
5%, and 1%, respectively.
VARIABLES
Network measures
Het_demographic
(1)
TobinQ
(2)
TobinQ
(3)
TobinQ
(4)
TobinQ
0.023
(0.184)
Het_intellectual
0.149**
(2.187)
Het_profession
0.128***
(2.880)
Het_international
0.467***
(3.364)
Het_overall
Centrality
Network size
Firm and CEO characteristics
Log(assets)
Leverage
Capextoasset
Cashflow
R&D intensity
Board female ratio
Tenure
Chairman
(5)
TobinQ
337.041***
(4.512)
0.021
(0.983)
292.105***
(3.910)
0.024
(1.146)
307.231***
(4.112)
0.003
(0.163)
344.239***
(4.634)
0.001
(0.024)
0.532***
(3.942)
297.094***
(3.851)
-0.005
(-0.190)
-0.140***
(-8.814)
-1.029***
(-8.721)
1.579***
(3.206)
0.524**
(2.404)
0.051***
(3.648)
0.638***
(2.826)
0.238*
(1.656)
0.006
(0.448)
-0.031
(-0.799)
-0.140***
(-9.155)
-0.988***
(-9.162)
1.529***
(3.482)
0.260
(1.348)
0.020**
(1.973)
0.725***
(3.314)
0.304**
(2.162)
-0.002
(-0.193)
-0.059*
(-1.677)
-0.132***
(-8.262)
-1.037***
(-9.211)
1.877***
(3.890)
0.176
(0.844)
0.023**
(2.017)
0.686***
(2.996)
0.216
(1.497)
-0.000
(-0.007)
-0.041
(-1.117)
-0.141***
(-8.847)
-1.025***
(-8.711)
1.611***
(3.272)
0.518**
(2.378)
0.051***
(3.653)
0.614***
(2.717)
0.228
(1.590)
0.006
(0.469)
-0.030
(-0.785)
-0.146***
(-8.624)
-0.939***
(-7.889)
1.632***
(3.099)
0.588**
(2.530)
0.049***
(3.207)
0.630***
(2.607)
0.276*
(1.813)
-0.003
(-0.260)
-0.046
(-1.162)
46
CEO Female
-0.143
(-1.005)
-0.008
(-0.684)
-0.133
(-0.935)
-0.015
(-1.396)
-0.120
(-0.863)
-0.013
(-1.031)
-0.140
(-0.978)
-0.008
(-0.634)
-0.136
(-0.843)
-0.017
(-1.401)
0.450
0.371
0.114
0.434
0.136
(1.187)
(1.003)
(0.412)
(1.137)
(0.442)
-0.108**
-0.109***
-0.104**
-0.105**
-0.095**
(-2.539)
(-2.755)
(-2.502)
(-2.463)
(-2.100)
3.783***
4.010***
3.850***
3.124***
3.736***
(4.362)
(5.233)
(4.270)
(3.484)
(4.183)
Year dummies
Yes
Yes
Yes
Yes
Yes
Industry dummies
Yes
Yes
Yes
Yes
Yes
Observations
10,272
11,336
10,919
10,248
8,846
Adjusted R-squared
0.136
0.135
0.140
0.139
0.140
CEO_age
Minority
Minority_missing
Constant
47
Table 6: Effects of CEO SNH on Firm Value: Economic Interpretations
This table reports the economic interpretations of the results in Table 5. Column (2) reports the coefficients of
estimates of CEO SNH on firm value. We also include the coefficient of centrality to compare with the
heterogeneity measures. The coefficient of centrality is taken from the regression on overall heterogeneity (Column
5 of Table 5). Column (3) reports the increase of market value if a CEO could increase his/her network
heterogeneity by 10%. Column (4) reports the calculation procedures.
Coefficients
Dollar amount (mil)
increase as a result of
10% increase in CEO
SHN
0.023
3.476
[10% * Mean (Het-demographic) * β (Hetdemographic) / Mean (Q)]*Assets =
(0.1*1.112*0.023/1.961)*2665.4=3.476
Het-intellectual
0.149**
28.231
[10% * Mean (Het-intellectual) * β (Hetintellectual) / Mean (Q)]*Assets =
(0.1*1.394*0.149/1.961)*2665.4=28.231
Het-profession
0.128***
29.681
[10% * Mean (Het-profession) * β (Hetprofessional) / Mean (Q)]*Assets =
(0.1*1.706*0.128/1.961)*2665.4=29.681
Het-international
0.467***
95.53
[10% * Mean (Het-international) * β (Hetinternational) / Mean (Q)]*Assets =
(0.1*1.505*0.467/1.961)*2665.4=95.53
Het-overall
0.532***
106.078
[10% * Mean (Het-overall) * β (Het-overall) /
Mean (Q)]*Assets = (0.1*1.467*
0.532/1.961)*2665.4=106.078
Centrality
297.094***
16.152
[10% * Mean (Centrality) * β (Centrality) /
Mean (Q)]*Assets =
(0.1*0.0004*297.094/1.961)*2665.4=16.152
CEO SNH
Het-demographic
Calculation
48
Table 7: CEO SNH and Firm Value: IV Approach
This table reports the results of instrumental variable regressions of CEO SNH on firm value (equation (3) and (4) as given in the text). Instrumental variable is
percentage of individuals who have network ties with the testing CEO and have either died or retired during the testing year. Column (1) and (2) report results of
the estimations relating the effect of CEO demographic heterogeneity on firm value. Column (3) and (4) report results of 2SLS estimations relating the effect of
intellectual heterogeneity on firm value using the same instrument. Column (5) and (6) report results for profession heterogeneity. Column (7) and (8) report
results for geography heterogeneity. Column (9) and (10) report results for overall heterogeneity. Control variables include CEO social network size and
centrality, key firm financial variables, board diversity, and CEO characteristics. Firm financial variables are measured at the previous year before the testing
year. Detailed variable definitions are available in Table 2. Year dummies and industry indicators at 1-digit SIC code are included. Numbers in the parentheses
are robust t-statistics. *, **, and *** denote significance at 10%, 5%, and 1%, respectively.
VARIABLES
(1)
(2)
(3)
(4)
(5)
(6)
(7)
(8)
Het - demo
TobinQ
Het - intell
TobinQ
Het - prof
TobinQ
Het - inter
TobinQ
Instrumental variable
Num_retire_death
-0.012**
(-1.986)
Network
measuares
Het_demographic
-0.094***
(-7.625)
-0.174***
(-10.208)
-0.016**
(-2.468)
(9)
Het overall
-0.069***
(-9.422)
0.968*
(1.827)
0.590**
(2.139)
Het_profession
7.744*
(1.912)
Het_international
1.298*
(1.736)
Het_overall
Network size
TobinQ
10.767*
(1.650)
Het_intellectual
Centrality
(10)
-36.441***
(-7.556)
0.068***
(52.610)
731.509***
(2.860)
-0.703
(-1.599)
123.075***
(11.710)
0.115***
(36.084)
193.002**
(2.341)
-0.066
(-1.118)
68.442***
(5.062)
0.167***
(42.423)
277.444***
(4.865)
-0.069
(-1.543)
-20.070***
(-4.537)
0.047***
(39.309)
492.951***
(4.549)
-0.336*
(-1.792)
35.727***
(6.882)
0.075***
(38.966)
270.679***
(4.377)
-0.059
(-1.088)
49
Firm and CEO characteristics
Log(assets)
0.001
Leverage
(0.995)
-0.014*
Capextoasset
(-1.762)
-0.037
Cashflow
(-1.216)
-0.012
R&D intensity
(-1.055)
0.003**
Board female
ratio
(2.044)
Board minority
ratio
Tenure
Chairman
CEO Female
CEO_age
Minority
Minority_missing
Constant
Year dummies
-0.152***
(-8.878)
-0.883***
(-6.279)
1.982***
(3.907)
0.653***
(3.267)
0.024
0.028***
(14.763)
-0.028**
(-2.091)
-0.095*
(-1.654)
-0.152***
(-8.989)
0.000
-0.163***
(-9.276)
-0.966***
(-14.282)
1.606***
(5.633)
0.386**
(2.491)
0.020**
0.015***
(5.193)
-0.047**
(-2.292)
0.045
(0.498)
-0.148***
(-5.441)
-0.000
-0.139***
(-13.702)
-1.017***
(-14.750)
1.854***
(6.059)
0.246*
(1.710)
0.023**
0.003**
(2.451)
-0.019**
(-2.498)
-0.058*
(-1.822)
0.001
(0.146)
0.001*
-0.160***
(-9.728)
-0.888***
(-7.697)
2.033***
(4.614)
0.510***
(3.235)
0.043***
0.011***
(9.038)
-0.036***
(-4.242)
0.004
(0.107)
-0.071***
(-5.907)
0.002**
-0.155***
(-11.691)
-0.912***
(-11.650)
1.628***
(4.966)
0.643***
(3.992)
0.047***
(1.037)
(0.334)
(1.986)
(-0.146)
(2.096)
(1.695)
(3.509)
(2.572)
(3.404)
0.051***
0.087
0.024
0.703***
-0.158***
0.757***
0.053***
0.222
-0.008
0.634***
(2.992)
0.101***
(8.555)
0.001
(1.455)
0.001
(0.221)
0.019**
(2.174)
-0.000
(-0.287)
0.023
(1.059)
-0.006*
(-1.928)
1.005***
(17.214)
Yes
(0.217)
-0.843
(-1.256)
-0.007
(-0.521)
-0.038
(-0.953)
-0.351*
(-1.956)
-0.006
(-0.653)
0.206
(0.545)
-0.045
(-0.796)
-7.034
(-1.064)
Yes
(0.874)
0.020
(1.192)
0.004**
(2.536)
0.008
(1.527)
0.037**
(2.374)
0.003**
(2.457)
0.033
(1.169)
0.006
(1.273)
0.605***
(6.265)
Yes
(5.206)
0.287***
(3.530)
-0.006
(-0.783)
-0.065***
(-2.838)
-0.164*
(-1.956)
-0.017***
(-2.772)
0.343
(1.530)
-0.114***
(-4.932)
3.511***
(6.605)
Yes
(-3.832)
-0.016
(-0.560)
-0.006***
(-3.413)
-0.026***
(-3.328)
0.007
(0.312)
-0.007***
(-4.276)
0.203***
(3.333)
0.011
(1.370)
1.526***
(12.788)
Yes
(5.300)
0.223***
(2.753)
0.003
(0.414)
-0.030
(-1.257)
-0.124
(-1.568)
-0.010*
(-1.692)
0.020
(0.119)
-0.109***
(-4.590)
3.137***
(5.363)
Yes
(3.378)
0.011
(1.101)
-0.001
(-0.847)
-0.001
(-0.408)
0.006
(0.714)
-0.001
(-1.372)
0.037**
(2.179)
-0.006**
(-2.126)
1.445***
(28.180)
Yes
(0.791)
0.152
(1.358)
0.011
(1.212)
-0.022
(-0.693)
-0.185*
(-1.735)
-0.001
(-0.091)
0.163
(0.519)
-0.060
(-1.435)
-7.408
(-1.259)
Yes
(-0.445)
0.032***
(2.893)
-0.002***
(-2.591)
-0.004
(-1.252)
0.024***
(2.735)
-0.002***
(-3.571)
0.085***
(3.513)
0.005*
(1.651)
1.325***
(23.384)
Yes
(4.372)
0.251***
(2.853)
-0.002
(-0.268)
-0.043*
(-1.730)
-0.155*
(-1.706)
-0.015**
(-2.257)
0.071
(0.375)
-0.099***
(-3.796)
2.714**
(2.486)
Yes
50
Industry dummies
F-stat from 1st
stage
Observations
Adjusted Rsquared
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
136.91
10,272
10,272
148.14
11,336
11,336
90.49
10,919
10,919
69.06
10,248
10,248
8,846
8,846
0.232
-1.402
0.319
0.105
0.208
0.119
0.129
-0.516
0.281
0.132
95.95
51
Table 8: Investor Response to CEO Appointment Announcement
This table presents the results of comparisons between two groups of firms that experience CEO turnover
events. Group 1 consists of firms where the new hire has greater social network heterogeneity than the old
hire. And Group 2 consists of firms where the new hire has the same or less social network heterogeneity
than the old hire. The two groups are formed using propensity score matching techniques to make sure two
groups of firms have similar firm characteristics including size, leverage, capital expenditure, cash flow,
and R&D intensity. We also require the new CEOs between two groups are similar in terms of age, tenure,
chairman position, and whether they are internal candidate or hired from outside. Panel A reports the
differences and associated t-statistics of firm and CEO characteristics between the two groups. Panel B
reports the results of the event study on new CEO appointment. We report cumulative abnormal returns
(CAR) over the 3-day window [-1, +1], where day 0 is the date on which the firm announces the new CEO
appointment. As alternative, we also compute 5-day window of [-2,2], and 10-day window of [-5,5].
Panel A. Matching on firm characteristics and CEO characteristics
Group 1: New CEO's
network heterogeneity
is greater than the old
CEO
Group 2: New CEO's
network heterogeneity is
equal or less than the
old CEO
Differences
P-value
Firm characteristics
Log(total assets)
7.392
7.558
-0.166
-0.522
Book leverage
0.167
0.184
-0.018
-0.565
Capex to assets
0.036
0.043
-0.007
-0.261
Cash flow
0.171
0.176
-0.003
-0.762
R&D intensity
0.068
0.055
0.013
0.415
CEO characteristics
CEO age
52.145
53.236
-1.091
-0.398
Tenure
13.524
12.433
1.091
0.399
Chairman
0.400
0.364
0.036
0.698
Hiring from outside
0.582
0.582
0.000
1.000
Variable Name
Panel B: Comparison of cumulative announcement return
CAR (-1,1)
0.014
-0.005
0.019*
0.093
CAR (-2,2)
0.018*
-0.011
0.029**
0.034
CAR (-5,5)
0.031**
0.0005
0.031*
0.092
52
Table 9: Regression Results Relating CEO SNH, Innovation, and Firm Value
This table reports the results of simultaneous equations estimations relating CEO social network heterogeneity,
innovation, and firm value. Innovation is measured by logarithm of patent applications. Column (1) and (2) examine
overall heterogeneity. To obtain an exogenous measure of overall heterogeneity, we use Het-overall-hat, which is the
predicted value of CEO overall social network heterogeneity from IV estimation using percentage of diseased or retired
social ties as instrument (results for the first stage estimation is available upon request). Column (3) and (4) examine
intellectual heterogeneity. We obtain Het-intellectual-hat from IV estimation using the same instrumental variable.
Column (5) and (6) examine professional heterogeneity. We obtain Het-professional-hat from IV estimation using the
same instrumental variable. Firm characteristics include firm size, leverage, cash flow, R&D intensity, capital
expenditure, indicator of high tech, indicator of multinational firms, and board diversity (female ratio and minority
ratio). We also control for network size and network centrality to tease out their potential effects on innovation and firm
value. Detailed variable definitions are available in Table 2. Year dummies and industry indicators at 1-digit SIC code
are included. Robust standard errors are clustered by firm. Numbers in the parentheses are robust t-statistics. *, **, and
*** denote significance at 10%, 5%, and 1%, respectively.
MODEL
Het-overall-hat
Simultaneous equations on
overall heterogeneity,
innovation, and Tobin's Q
intellectual heterogeneity,
(1)
(2)
(3)
Log (patent)
Tobin's Q
Log (patent)
5.172***
1.778*
(5.052)
(1.702)
Het-intellectual-hat
(4)
Tobin's
Q
3.988***
1.445*
(5.011)
(1.699)
Het-profession-hat
Lag (patent)
professional heterogeneity,
(5)
(6)
Log (patent)
Tobin's Q
1.283***
0.718*
(2.821)
(1.727)
0.022**
0.023**
0.025**
(2.012)
(2.157)
(2.342)
Controls
Networksize
Centrality
Log(assets)
Leverage
Capextoasset
Cashflow
R&D intensity
High tech
-0.281***
-0.107
-0.330***
-0.131
-0.124*
-0.097
(-3.817)
(-1.437)
(-3.943)
(-1.483)
(-1.670)
(-1.436)
396.542***
132.613**
64.478
9.513
461.202***
141.167**
(5.549)
(2.137)
(0.577)
(6.580)
(2.351)
0.411***
-0.134***
0.368***
(0.086)
0.153***
0.474***
-0.119***
(18.740)
(-6.041)
(12.615)
-0.595***
-0.515***
(-5.489)
(-5.389)
-0.874
(29.446)
(-7.827)
-0.668***
(-4.879)
0.542***
-0.730***
-0.572***
(-6.298)
(-5.935)
(-6.906)
(-6.338)
2.690***
-0.305
2.990***
-0.973*
2.683***
(-1.488)
(5.688)
(-0.501)
(5.742)
(-1.652)
(5.681)
0.498***
1.505***
0.679***
1.578***
0.543***
1.540***
(3.500)
(12.766)
(4.631)
(12.733)
(3.797)
(12.945)
2.917***
3.801***
3.129***
3.879***
3.501***
3.965***
(10.252)
(13.466)
(11.884)
(15.419)
(14.045)
(17.756)
-0.564***
0.510***
-0.500***
0.527***
-0.401***
0.546***
53
(-6.198)
(6.307)
(-5.891)
(6.971)
(-4.861)
(7.783)
0.084
0.969***
0.163
1.003***
0.320
1.101***
(0.412)
(5.560)
(0.805)
(5.763)
(1.513)
(5.938)
0.440***
-0.174
0.504***
-0.150
0.635***
-0.102
(3.457)
(-1.557)
(4.067)
(-1.398)
(5.234)
(-0.980)
Constant
-6.800***
0.000
-4.324***
0.000
0.000
1.544**
(-5.384)
(.)
(-5.243)
(.)
(.)
(2.471)
Yes
Yes
Yes
Yes
Yes
Yes
Board female ratio
Year dummies
Industry dummies
Yes
Yes
Yes
Yes
Yes
Yes
Observations
6,160
6,160
6,160
6,160
6,160
6,160
Adjusted R-squared
0.395
0.214
0.395
0.214
0.393
0.214
54
Table 10: Regression Results Relating CEO SNH, Foreign Sale Growth, and Firm Value
This table reports the results of simultaneous equations estimations relating CEO social network heterogeneity, foreign sale
growth, and firm value. Foreign sale growth is measured by percentage change of foreign revenue from last year. Column (1)
and (2) examine overall heterogeneity. To obtain an exogenous measure of overall heterogeneity, we use Het-overall-hat,
which is the predicted value of CEO overall social network heterogeneity from IV estimation using percentage of diseased or
retired social ties as instrument (results for the first stage estimation is available upon request). Column (3) and (4) examine
demographic heterogeneity. We obtain Het-demographic-hat from IV estimation using the same instrumental variable. Column
(5) and (6) examine international heterogeneity. We obtain Het-international-hat from IV estimation using the same
instrumental variable. Firm characteristics include firm size, leverage, cash flow, R&D intensity, capital expenditure, indicator
of high tech, indicator of multinational firms, and board diversity (female ratio and minority ratio). We also control for network
size and network centrality to tease out their potential effects on foreign sale and firm value. Detailed variable definitions are
available in Table 2. Year dummies and industry indicators at 1-digit SIC code are included. Robust standard errors are
clustered by firm. Numbers in the parentheses are robust t-statistics. *, **, and *** denote significance at 10%, 5%, and 1%,
respectively.
MODEL
VARIABLES
Het-overall-hat
overall heterogeneity, foreign
sale growth, and Tobin's Q
(1)
Foreign sale
growth
(2)
Tobin's Q
2.013*
1.195
(1.826)
(1.242)
Het-intellectual-hat
intellectual heterogeneity, foreign
(3)
Foreign sale
growth
(4)
Tobin's Q
8.718*
5.176
(1.826)
(1.242)
Het-profession-hat
Lag(Foreign sale
growth)
professional heterogeneity, foreign
(5)
Foreign sale
growth
(6)
Tobin's Q
4.544*
2.698
(1.826)
(1.242)
0.082***
0.082***
0.082***
(6.662)
(6.662)
(6.662)
Controls
Networksize
-0.115*
-0.045
-0.531*
-0.292
-0.195*
-0.093
(-1.723)
(-0.780)
(-1.810)
(-1.142)
(-1.774)
(-0.969)
Centrality
-107.589*
126.092**
189.802
302.666**
53.054
221.471***
(-1.778)
(2.387)
(1.362)
(2.488)
(0.710)
(3.395)
Log(assets)
-0.054**
-0.030
-0.062**
-0.035
-0.054**
-0.030
(-2.512)
(-1.600)
(-2.456)
(-1.579)
(-2.512)
(-1.601)
-0.091
-0.882***
-0.047
-0.856***
-0.073
-0.871***
(-0.941)
(-10.461)
(-0.480)
(-10.022)
(-0.756)
(-10.356)
-0.045
-2.262***
-0.049
-2.264***
0.087
-2.184***
(-0.097)
(-5.606)
(-0.105)
(-5.613)
(0.185)
(-5.296)
-0.871***
5.469***
-1.329***
5.197***
-1.163***
5.296***
(-4.828)
(34.648)
(-4.531)
(20.261)
(-5.096)
(26.534)
-1.175*
4.920***
-1.900*
4.490***
-0.862*
5.106***
Leverage
Capextoasset
Cashflow
R&D intensity
55
High tech
Board female ratio
Board minority
ratio
Constant
Year dummies
Industry dummies
(SIC1)
Observations
Adjusted Rsquared
(-1.837)
(8.811)
(-1.881)
(5.094)
(-1.750)
(11.878)
0.001
0.477***
0.287*
0.647***
0.002
0.478***
(0.016)
(7.175)
(1.888)
(4.873)
(0.031)
(7.203)
-0.353**
0.540***
-0.830***
0.256
-0.629***
0.376**
(-2.097)
(3.679)
(-2.814)
(0.994)
(-2.946)
(2.016)
0.103
0.201**
-0.886
-0.386
0.092
0.195**
(0.963)
(2.149)
(-1.497)
(-0.747)
(0.838)
(2.030)
-1.185
0.287
-7.116
-3.235
-4.678
-1.787
(-0.975)
(0.270)
(-1.602)
(-0.835)
(-1.505)
(-0.659)
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
5,037
5,037
5,037
5,037
5,037
5,037
0.037
0.332
0.037
0.332
0.037
0.332
56
Table 11: Regression Results Relating CEO SNH, Investment Efficiency, and Firm Value
This table reports regression results relating equation (9) as given in the text. Column (1) shows the estimation of
investment sensitivities to Tobin’s Q using total capital expenditure ratio as the dependent variable. Column (2) adds
the interaction terms of CEO social network heterogeneity with Tobin’s Q. The coefficients of the interaction terms
capture how investment sensitivity to Q changes as CEO social network heterogeneity increases. A positive
coefficient of the interaction term means that investment-Q sensitivity increases as CEO network heterogeneity
increases, and an increase in investment-Q sensitivity indicates improved investment efficiency. To obtain an
exogenous measure overall heterogeneity, we use Het-overall-hat, which is the predicted value of CEO overall
social network heterogeneity from first stage of IV estimation using percentage of diseased or retired social ties as
instrument. Column (3) and (4) repeat the same analysis using acquisition expenditure to asset as dependent variable
to capture the efficiency of acquisition related investment. Detailed variable definitions are available in Table 1. All
models include firm fixed effects. Numbers in the parentheses are robust t-statistics. *, **, and *** denote
significance at 10%, 5%, and 1%, respectively.
VARIABLES
Het-overall-hat
(1)
total capital
expenditure ratio
Het-overall-hat *Tobin's Q (t-1)
Tobin's Q (t-1)
Inverse logasset (t-1)
Leverage (t-1)
Cash flow (t-1)
Constant
Firm fixed effect
Observations
Number of firms
Adjusted R-squared
0.021***
(11.368)
1.884***
(6.733)
-0.166***
(-10.258)
0.017
(0.873)
0.137***
(15.976)
Yes
9,788
2,375
0.086
(2)
total capital
expenditure ratio
1.088
(1.410)
0.299***
(2.739)
0.022***
(8.769)
1.863***
(4.022)
-0.188***
(-8.640)
0.001
(0.031)
0.131***
(9.575)
Yes
7,820
2,099
0.078
(3)
acquisition
expenditure ratio
`
0.026***
(3.693)
5.490**
(2.083)
-0.385***
(-9.178)
0.274***
(3.529)
0.064**
(2.243)
Yes
6,386
1,814
0.098
(4)
acquisition
expenditure ratio
2.231
(1.534)
1.043***
(3.529)
0.042***
(4.731)
2.481
(0.861)
-0.438***
(-7.967)
0.174*
(1.928)
0.089***
(2.848)
Yes
5,265
1,612
0.098
57
Table 12: CEO SNH and M&A performance
This table reports the OLS regressions results relating CEO social network heterogeneity and M&A
performance. Dependent variable in Column (1) and (2) is short run performance, measured by cumulative
abnormal returns around [-2,2] of the announcement date. Dependent variable in Column (3) and (4) are
long run performance, measured by BHAR (buy and hold abnormal return) during 3-year time period post
M&A. BHAR is estimated using value weighted market returns as benchmark. Independent variables of
main interest are overall CEO social heterogeneity and its interaction with diversified M&A. We control
deal characteristics and acquirer financial variables. Detailed definitions are provided in Table 1. Numbers
in the parentheses are robust t-statistics. *, **, and *** denote significance at 10%, 5%, and 1%,
respectively.
VARIABLES
Het-overall-hat
(1)
CAR[-2,2]
0.298**
(1.960)
(2)
CAR[-2,2]
0.266**
(1.964)
0.009**
(2.401)
(3)
BHAR- 3year
4.461*
(1.735)
(4)
BHAR- 3year
5.205*
(1.766)
1.963**
(2.197)
-0.020**
(-2.041)
0.003
(0.445)
-0.006
(-0.986)
-0.030***
(-3.432)
0.019
(1.533)
-0.476
(-0.036)
-0.019**
(-2.103)
0.003
(0.521)
-0.004
(-0.743)
-0.028***
(-3.657)
0.018*
(1.768)
-2.586
(-0.219)
-0.018***
(-3.040)
0.003
(0.026)
-0.023
(-0.362)
-0.062
(-0.979)
-0.215**
(-2.152)
0.142
(1.145)
331.354**
(2.402)
0.008
(0.071)
-0.026
(-0.383)
-0.074
(-1.045)
-0.238**
(-2.122)
0.169
(1.236)
345.398**
(2.266)
-3.021**
(-2.266)
-0.013**
(-2.362)
0.063***
(3.013)
-0.003
(-1.011)
0.018
(1.008)
-0.008
(-0.850)
0.104***
(2.591)
0.041
(1.245)
0.003
-0.012**
(-2.373)
0.062***
(3.489)
-0.003
(-0.939)
0.018
(1.070)
-0.009
(-1.029)
0.092**
(2.536)
0.042
(1.381)
0.007
-0.063
(-1.276)
0.506**
(2.115)
-0.048
(-1.466)
0.007
(0.041)
-0.100
(-1.280)
0.790
(1.509)
-0.739**
(-2.059)
-0.123
-0.096
(-1.516)
0.586**
(2.134)
-0.062*
(-1.645)
-0.031
(-0.173)
-0.064
(-0.739)
0.992
(1.626)
-0.788*
(-1.935)
-0.217
Het-overall-hat* Diversifying M&A
Deal characteristics
All stock payment
Mix cash and stock payment
Private target
Public target
Tender offer
Relative size
Diversifying M&A
Acquirer characteristics
Log (assets)
Leverage
Market to book
Run-up
Sale growth
ROA
Board female ratio
58
Constant
Year dummies
Industry dummies
Observations
Adjusted R-squared
(0.147)
-0.393**
(-2.030)
Yes
Yes
3,757
0.023
(0.342)
-0.353**
(-2.194)
Yes
Yes
3,781
0.059
(-0.467)
-5.720*
(-1.920)
Yes
Yes
3,781
0.060
(-0.723)
-6.299*
(-1.891)
Yes
Yes
3,757
0.023
59

Heterogeneity of CEO Social Networks and Firm Value

Transcription

Similar documents

Growco Thursday AM Gen Session.pptx

Fact Sheet - Global Retailing Conference

on Mark Scharenbroich and his Nice Bike message . . .

Team Engagement

Market-at-a-glance - Smart Business Magazine

Sunday,Sept.18,2016 - Carz Cruizin to Cure Cancer

lpaa focus on private finance

Diapositiva 1

UIBS 2013 Flyer

An open letter to Prime Minister Stephen Harper, Thomas